Reviews & Opinions
Independent and trusted. Read before buy Doro Sensor 100!

Doro Sensor 100 Manual

Preview of first few manual pages (at low quality). Check before download. Click to enlarge.
Manual - 1 page  Manual - 2 page  Manual - 3 page 

Download (Swedish)
Doro Sensor 100 - Hurtig Guide, size: 356 KB
Instruction: After click Download and complete offer, you will get access to list of direct links to websites where you can download this manual.


Doro Sensor 100About Doro Sensor 100
Here you can find all about Doro Sensor 100 like manual and other informations. For example: review.

Doro Sensor 100 manual (user guide) is ready to download for free.

On the bottom of page users can write a review. If you own a Doro Sensor 100 please write about it to help other people.
[ Report abuse or wrong photo | Share your Doro Sensor 100 photo ]

User reviews and opinions

No opinions have been provided. Be the first and add a new opinion/review.




New Zealand

Dedicated Network services telephone

User guide

The SENSOR 500 is a dedicated network services telephone. SENSOR 500 is compatible with Caller ID, Call Answer and message waiting network services. In addition it is also pre-programmed for access to other optional services. Up to 10-speed dial memories and 20 name and number phonebook add to its versatility Other features include: headset socket, on-hook dialling and a choice of 12 languages for the display text. Note: Caller display, Call Answer and other optional services are usually subscription services.

Installation.. 3

Location of controls.. 3 Connection.. 4 Wall mounting. 5 Headset..5 Select language. 5 Display.6 Set date and time.6 Using the telephone. 7 Making a call. 7 On the hook dialing.7 Using a headset. 7 Last number redial.. 8 Volume control.. 8 Mute.. 8 Pause.. 8 Telephone ringer.. 8
Caller display.. 9 New number(s).. 9 Reviewing the call list.10 List information..10 Erase phone number(s). 11 If the call list is full. 11 Dialling from the call list. 11 Speed dial memories.. 12 Store number/name.. 12 Phonebook..13 Store..14 Erase.. 14 Dial... 14 Call Answering..15 Miscellaneous..16 Trouble shooting.. 17 Guarantee.18


Handset Curly cord socket Wall mounting hook Hook switch Dial-Up key Scroll call list/phonebook, replay Line cord socket (rear) Display Message waiting indicator Play/Pause messages Headset socket (rear) Switch between speaker/headset Switch for ringer level
Erase key Scroll call list/phonebook, save Speed-dial/pre-programmed keys Volume control for handset Volume control for speaker Mute key Key/indicator for speaker/headset Dial keys 0-9, *, # Programming key Redial key Recall-key Phonebook access
Connect the curly cord to the left-hand side of the telephone and to the bottom of the handset. Connet the power cord to middle socket of telephone and plug into power outlet. Connect the telephone line cord to the socket entitled LINE on the rear of the telephone and then connect the cord to the network wall outlet. The display will show: NO CALLS.
Headset socket Line cord socket
Plug pack power To connect headset, see page 4. Do not place the unit in direct sunlight or in an area of high humidity. In very rare situations the rubber feet on the base of the telephone may discolour a surface. It is recommended that you disconnect the base unit from the mains power during a thunderstorm. Do not use in the rain, near a bath or with wet hands. To clean the product, use a damp cloth or antistatic wipe. Never spray cleaner fluid directly onto the handset or base unit.


The phone can be wall mounted, using the bracket supplied. Use the bracket supplied to mark the position of screw holes on the wall. Drill the holes (check for hidden pipes, etc). Use the raw plug and screws, ensuring approximately 3 mm remain between the wall and the head of the screw. Connect the bracket to the rear of the base unit and place the complete assembly above the screws and gently push it down onto the screws. Rotate the handset hanger hook 180 degrees.


A headset can be connected to the socket telephone. located on the rear of the
1. Connect headset to headset socket at the rear of the phone Headset socket
2. Select headset position, using the switch on the right hand side of the telephone. On-hook dialling: Headset: NOTE! We recommend that you use a Plantronics headset.


12 languages are available for the dislpay texts: English, Swedish, Finnish, Danish, Norwegian, German, Spanish, Dutch, Portuguese, Turkish, Italian and French. 1. Press down 2. Press or until the display shows - - - - - - - -, then release. to select LANGUAGE. or.

3. Press and release

4. Select a language, using 5. Press and release

to store the setting.

When the unit is in standby mode the display will show: the current date and time, if new calls have been received (if these have not been reviewed or answered) and the total number of calls on the call list. Text information


:88 88/18
No./ Time Position Date New calls


The current date and time will be set automatically once a call is received. However, if you want to set these values yourself, follow the procedure below: Press and hold Press and release Press and release. or. until the display shows - - - - - - - -., then release. , until the display shows SET TIME.
The hour is now flashing on the display. Adjust with Press C to store the setting and continue. The minute is now flashing on the display. Adjust with Press C to store the setting and continue. The day is now flashing on the display. Adjust with Press C to store the setting and continue. The month is now flashing on the display. Adjust with Press to store the setting and return to stand-by


A call can be made in several different ways, using the handset, on-hook dialling or headset.


1. Lift the handset and wait for dial tone. 2. Dial the phone number required. To terminate the call replace the handset. You can change to speaker/headset during a call by pressing and replacing the handset. Note: Ensure the switch on the right hand side is set to the desired function headset/speaker.
1. Ensure that the switch on the right hand side of the phone is set to. 2. Press and release. Use the volume control on the right hand side to adjust the sound level. 3. Dial the number required. 4. When the called party answers, lift the handset. To terminate the call replace the handset.


1. Ensure that the switch on the right side of the phone is set to: 2. Press and release. Dial tone is heard in the headset.
3. Dial the number required. Press and release to terminate the call.


If the last number dialled was engaged or the call was not answered, it is possible to redial the number using the button. Press and release. The telephone number will be dialled automatically.
When the called party answers, lift the handset (or headset) to converse. If the call is not answered, press and release Its also possible to lift the handset and then press the redial key.


The sound volume in the handset or headset can be adjusted using the [VOL ] to increase the volume , and [VOL ] to decrease the volume.
If during a telephone conversation, you wish to talk privately to a third party without the caller overhearing: Press the button (the word MUTE will appear on the display). To resume the conversation, press the button again.
A short pause to numbers stored in the phone book and the speed-dial memories, using this button. This may be required if the phone is connected to a PABX or if used with some network services.


The telephone ringer volume can be set to either high, low or off which is represented by this symbol.


Caller display allows you to see who is calling before you answer a call and to see who has called in your absense. When the service is implemented on your telephone line, seconds before your telephone rings the number of the incomming caller will appear on the display. Up to 60 incomming numbers can be stored in the call list. When the list is full, the oldest number will be deleted automatically. It is not possible to view the call list and take a new call at the same time. NOTE! Caller Display is an optional telephone Network service. If the caller Display service is not implemented on your telephone line, incomming call information will not be displayed.


The words NEW CALL and the number of calls will appear on the display. The envelope symbol on the display will flash. Once the number(s) have been reviewed, NEW CALL will disappear and the indicator will stop flashing. When an incoming number is received it is assigned a call sequence number which appears on the top left hand side of the display, commencing with 01 (oldest call) up to 60 (latest call).


Scroll the memory with the keys or. The key Scrolls from the oldest number. The key Scrolls from the latest number. The time and date each call is received is shown on the display. If the number is stored iin the phonebook the corresponding name is also shown.

9 - 2805000

The display will also provide the following information:

:56 14/10

Phone number 9-2805000, received at 10:56 the 14th October. Position number 17.


You may erase a single number or all numbers on the call list.


Scroll the call list until the number to be erased is shown on the display. Press C twice quickly, then release


Scroll the call list until a number is shown on the display. Press and hold C until the display shows ERASE ALL?., then release. Press and release C again. All phone numbers are now erased.


The call list memory capacity is 60 telephone numbers. When the memory is full, each new coming call will replace the oldest stored number, the principle first in first out applies. The new number receives the highest number, 60, and all other numbers are moved one step up in the memory order.
Scroll the call list with displayed. or until the desired phone number is. and then lift the handset when
Lift the handset and press and release It is also possible to first press and release the called party answers. NOTE!
If a charge for local calls is unacceptable, the Dial Button should not be used for local calls when using some Service providers. Only the 7-digits of the local number should be dialled from your telephone. Do not dial the area code or the 0 prefix


Ten One-touch speed dial memories are available. Memories 1 & 2, 6 & 7 have been pre-programmed for easy access to and set of for select services. If required these memories can be re-programmed with numbers of your own choice following the procedure below.
Memroy 1 = Call waiting answer / Call hold Memory 2 = Stop your telephone number being displayed for this call Memory 6 = Program call foward always Memory 7 = Turn Call foward always off


1. Press. 2. Dial the phone number to wish to store (16 digits maximum). 3. Press and release.
4. Enter a (15 letters maximum) by pressing the relevant keypad digit buttons once or several times (see page 13) - If the same letter is you want to go to the next position press. - For edit, press. - To make a space, press 0. 5. Press and release the required speed-dial key to store the entry. If you want to erase a speed-dial number, replace it with a new number.


1. Press and release the desired speed-dial key. The number is dialled automatically. 2. Lift the handset when the other person answers (or use a headset). Its also possible to lift the handset first and then press the speed-dial key.


When any of the numbers that are stored in the one touch memories call you, an alternative ringer tone is heard, this feature allows you to distinguish between callers, without looking at the display. 12
The telephone has a phone book where you can store 20 names and telephone numbers for easy access and fast dialling. When one of these phone numbers calls, both name and number will be displayed. In stand-by display will alternate between NEW CALLS and VIP on the upper line. This ensures that you see when a call that matches a number in the phone book (or speed-dial memory) is received.
20 call/reviewing 16/:08 Incoming


22 20::08 18/10 Stand-by
The keypad digit buttons are also used to store letters in the phone book. Printed on each keypad button are a number of letters which in total make up the english alphabet. Other letters and some special symbols for example ! and & exist in the memory even though they are not printed on the buttons, the table below indicates where you will find them.
1. Press and release. The display shows how many numbers/ names are stored in the phonebook. 2. Press.
3. Dial the full number to be stored as you would manually dial that number (16 digits maximum). 4. Press and release.
5. Enter a name. (15 letters maximum) by pressing the relevant keypad digit buttons once or several times (see page 13) - If you have to enter the same letter twice, press and release onto the next position - For edit, press. - To make a space, press 0. 6. Press and release. to move
Continue to store numbers from point 1 above or press and release to exit.
Press , the display shows the amount of stored numbers in the phonebook. Use or to scroll the desired number.
Press C twice to erase the number. Continue to erase numbers or press to return to standby mode.
1. Press , the display shows the amount of stored numbers in the phonebook. 2. Select the number required with 3. Press and release. 14 or.
4. Lift the handset when the called party answers (or talk in a headset).


To check your messages, simply press Play and your messages will be played to you. Whilst listening to your messages you can use the keys featured below to operate various options. Backwards Pause/Play Forward Erase current message To use the other features of the network service, press the corresponding key. Message waiting indicators The LED indicator will flash The display will alternate NEW CALLS and MSG WAITING These indicators will only work when you have subscribed to a network voice message service. (Telecom Call Minder) You will have to retrieve all your messages before they are extinguished. Note: These indicators will be reset to off when the phone is disconnected or during a power failure.

RE-PROGRAM FOR ALTERNATIVE CALL ANSWERING SERVICE If the pre-programmed keys are not compatible with your network operators answering service, it is possible to re-program the keys. Press. Press and hold down Enter digit (s) to store Press the key required. Press Press to store to finish. until the display shows text, then release


THE RED LED INDICATOR ON THE BASE UNIT IS FLASHING There is a message waiting in your Network voice message system.
Have you subscribed to the caller display service? If Out of area or Withheld Secret are displayed, it either means that no information of the number is available or that the caller has withheld their number. Is the phone connected to a PABX? Caller display is not available when connected to a PABX.
Check that the ringer is switched on. Does the total of RN numbers exceed 4? If it does, disconnect other equipment until 4 is achieved.


Is the telephone line cord plugged into a telephone network wall socket?
Store a number and try again.
The RN (Ringer Equivalance Number) is of significance only if you wish to connect more than 1 telephone to your telephone line. Your telephone line has a maximum RN capacity of 5. Your Audioline Sensor500 has a RN of 0.5, thus it is feasible to connect 4 more telephones of a value of 1 to your system. You should not exceed a value of 5 on the complete system, otherwise the volume of the ringer in any phone will decrease and one telephone may not ring at all.


If you believe your SENSOR 500 is malfunctioning, please refer to the relevant section and/or consult the troubleshooting guide in this manual to ensure that you have followed the instructions carefully. The SENSOR 500 is warranted for a period of 12 months from the date of purchase by you, the end user. In the unlikely event of a fault during this period. please contact our Helpline for assistance. If the product is then found to be faulty you will be asked to return it directly to the service agent, ATLAS GENTECH (NZ) LIMITED with a copy of the purchase receipt. ATLAS GENTECH (NZ) LIMITED 87 Carbine Road, Mt Wellington, Auckland Service & Spare Parts (Toll Call): 025 Fax: (09) 574-2722 Email: NOTE: The guarantee does not extend to damage caused by misuse, negligence, excessive voltage, faults on the telephone line or lighting. This guarantee in no way affects your statutory right. DORO is the trade mark of DORO AB.

The grant of a Telepermit for any item of terminal equipment indicates only that Telecom has accepted that the item complies with minimum condition for connection to its network. It indicates no endorsement of the product by Telecom, nor does it provide any sort of warranty. Above all, it provides no assurance that any item will work correctly in all respects with another item of Telepermitted equipment of a different make or model, nor does it imply that any product is compatible with all of Telecoms network services. Warning: This equipment may not provide for the effective hand-over of a call to another device on the same line. This equipment shall not be set to make automatic calls to the Telecom 111 Emergency Service.


DORO Technical Report

Andrea Carbone, Giorgio Ugazio, Alberto Finzi, Fiora Pirri ALCOR Laboratory Dipartimento di Informatica e Sistemistica Universit` di Roma La Sapienza Via Salaria 113, I-00198 Roma, Italy a { carbone,ugazio,finzi,pirri} May 6, 2004


In this rst report we are going to describe the state of the art of our hardware platform, his equipment and a sketch of the principal components of the proposed architecture. First we will briey step into the description of the hardware platform (in section: 2) then after a peek on the distributed software architecture that we are going to shape (in section: 3), we will introduce the heart of the system: the mission planner (in section: 5). A description of the low level modules (localization and mapping (sec: 6) and an introduction to the reactive layer (sec:7)) closes our dissertation.

The Main Character: DORO

DORO (represented in gure 1)is a two-wheeled dierential drive robot with a caster wheel on the rear. The basis is a Pioneer 3DX ActivMediaTM ( autonomous robot equipped with various devices necessary to: acquire and deliver perceptual snapshots of the surrounding environment; collect information about internal state in order to track and control the evolving state of the navigation inside the arena. The rst device set is mainly represented by a couple of CCD cameras mounted over a Pan Tilt Unit that let the underlying proprietary processes to drive the attention toward interesting regions of the perceived scene. On the same pan tilt a laser telemeter can retrieve in an asynchronous or synchronous fashion highly precise metrical measures of details tagged as interesting features from the online visual inference. The second set is composed by the embedded acoustic sonars and odometric sensors plus an additional and external inertial platform whose main purpose is to provide a more ne grained integration of the position and orientation in order to recovery the odometric errors due to slippage or low accuracy data coming from wheels.

Distributed System

Many processes share devices and consume system resources (cpu, memory, time). A distributed software system is a practical and advantageous solution to implement all activities and give shape to a exible and reliable system. All tasks are thinked as independent processes who listen for particular events to happen remotely and react in an opportune consistent way. So, in brief, a multithreaded networking environment 1
Figure 1: Our Rescue Robot that we named DORO with proxies object incapsulating activities along with distributed publish/subscribe policies [GV02] has been sketched to realize an extensible and robust rescue system.

Sharing resources

All devices (robot included) can be seen as system resources shared by the logical components that shapes the cognitive architecture of the overall system. While some device belong exclusively to a precise module, some other resources can be shared among more than one subsystem (i.e. the pan tilt unit). An harmonic access to shared resources is carried out by a custom module (described in his general issues and characteristics in section 5) whose duty (among many others) is to avoid and discipline potential conicts between multiple clients. We will see later that such module provides more than static rules for this coordination but a real reactive and adaptive task scheduler.
The Beating Heart of the System: the Mission Planner

General Aspects

Autonomous mobile mission planning supports the optimization of real-world missions involving multiple concurrent goals. Typical rescue mission goals are sets of scientic observations/experiments which have to be performed. For example, some mission level goals could be take N pictures in the forward direction

while doing a D degrees rotation on the site or Move to an unexplored region., Approach an interesting target in the D direction., etc. The Mission Planning is to squeeze in as many of the science goals as possible, check that the plan stays within safe boundaries for resources like battery power and enforces the science constraints like scheduling the pictures at sunrise. If, for example, the Mission Planner nds that executing the plan would drain the rovers battery midway, it has to rearrange the tasks in order to remove the violation (or remove some goals), regenerating a feasible plan. The Mission Planner is a crucial component of the autonomous system since it keeps a global (and abstract) representation of the rover state and goals and continuously adapts the execution of the mission tasks as perceptual information about the domain is acquired. For example, if an interesting object is detected while the rover is moving toward the site Y, the Mission Planner can decide to stop the robot and perform an observation by acquiring rights to use the Pan Tilt Unit and the camera. This decision can be taken only if the predicted eort does not violate the mission and rover hard constraints (e.g. the rover is to reach the site Y within a time windows). In order to perform prediction and choices for achieving mission goals and utilities, mission planning relies on: Declarative Models of the environment and of the robot capabilities (i.e. description of the tasks preconditions, eects, resources and time constraints); Specication of required goals and/or utility criteria; Online input from sensors and communication channels. The specics of planning in robotics are mainly the need to handle: Heterogeneous partial models of the environment and of the robot components, as well as noisy and partial state information acquired through sensors and communication channels Direct integration of planning to acting and sensors outcomes. The proactivity of the autonomous system refers, roughly, to the choice of an appropriate course of abstract actions based on the model of the world and the mission execution context (mission and robot state, sensory data). In contrast to decision making in the reactive layers, the mission planner considers more than the current local state, which derives largely from what is currently perceived, but takes into account background knowledge, projects various possible courses of action in a goal-directed fashion and then chooses one depending on the given goals or some measure of utility. 5.1.1 Model-based Autonomy

The declarative model is the key element of the so called model-based autonomous agent: given a formal specication of the domain, the agent can deploy deliberative activity (planning, constraint solvers etc.) in order to check the consistency of the robot behavior (with respect to the model) and decide how to achieve the mission goals. The model-based approach is to enhance the exibility and the safety of the robotic system: on the one hand the mission plans can be recongured on the y depending on the execution contexts, on the other hand, the rover activities can be continuously monitored with respect to a model, hence failure detection and recovery are simplied. This high level monitoring process (monitoring the system activities w.r.t. the declarative model of its behavior) is to reinforce and coordinate the low-level fault protection mechanisms. For example, if a low level process is interrupted by a low-level fault protection mechanism, this event is to be detected by the high level monitoring system which can decide to relay on/activate one of the spare low-level capabilities in order to put the system in a safe mode. The model-based monitor is aware about these capabilities (like the hazard avoidance functionalities) only if these are represented in the robots declarative model, in this case it can decide how to react to the failure by assessing such model and the execution context.
Planning and Execution Interaction
The Planning capabilities described above have to be closely integrated with the reactive one. This integration is a critical issue in an autonomous robot architecture, a tight integration can be obtained by combining the classical plan generation process with the plans execution: the execution components continuously report the planning system about the progress being made, including action failures and unexpected world changes. The continuous sensors feedback enables the planning system to adapt exible plans to the new execution contexts or to replans when it is necessary. Hence, dierently from to the Path Planning processes, the Mission Planning is to be synchronized with the low level reactive system: relevant low-level data (ended/interrupted/failed tasks, sensing data etc.) continuously updates the mission state so that the mission planner can check the validity of the current plan by assessing temporal and resource constraints. In this setting, since the deliberative (mission) planner has to take into account also the scheduling problems, a more detailed model of the control system is needed and the scope of the model based autonomy is extended to the executive system. The executive control system modules must be associated with a detailed temporal model of their activities so that the execution monitor can work as a reactive planner in order to schedule the parallel tasks on the timelines in tight cooperation with the mission planner. For example, we can assume a robot approaching a target point in order to be there within a time windows (e.g. between 30 and 60 seconds). If an interesting object is detected while the rover is moving, is up to the execution monitor to assess the time/resources available and to schedule a sequence of actions in order to observe it. For instance, the monitor could decide to stop the rover, approach the object and deploy the Pan Cam to take a spectral images of it, but this activity must be consistent with the long term goal and with the other parallel activities the rover is performing. A second possible plan consists of stopping the robot and taking a low resolution panorama of the area, performing on-board image analysis to nd rocks in the panorama. Another possibility is to just neglect the object. These alternatives can be assessed only if the scheduler is endowed with a detailed model of the components behavior (duration and cpu/energy cost of on-board processes, cost/risk associated to navigation etc.). For instance, if the rover is performing some heavy data processing, on-board image analysis could not be possible. 5.1.3 Learning capabilities

The interaction between the mission planner, the execution monitoring system, and the reactive system can be supported by probabilistic models and learning techniques formally interfaced with the representational framework. In this setting, time and resources decisions can be enhanced by deploying Bayesian analysis for resource allocation. This is based on utility and risk functions dened according to statistical models accounting for the underlying behaviors. Given a set of tests capturing the relevant states of request/error/failure, the Bayesian analysis can deploy learning and classication models in order to account for requests and commands synchronization, in so ensuring a system adaptability.
High Level Functionalities
In the following we shall indicate, with the aid of some schemata, the role of the mission planning activities together with the execution activities, integrated with the necessary learning activities. 5.2.1 Mission Planning and decision (MP)
The MP is devoted to a goal-driven management of parallel activities within compound tasks, under temporal constraints and predictable part of the changing environment, e.g., contingent change not under the robot control such as day/night cycles, resource availability proles, or expected events. The MP activity can be summarized as follows: Input. State of the agent: partial plan, execution state, and pending goals. Input. Temporal dynamic model of the robots modules and environment. Input: Local path planning and behaviors.
Activity. Long-medium term planning according to time constraints, resources. Output. A mission plan for the next goals: actions are processes indexed with time (i.e. start end+ constraint), positions are spatial-qualitative cells. Several techniques shall be used to cope with this issue, including symbolic formalisms such as the Temporal Situation Calculus and First Order logic, necessary to ensure a deliberative model for the activities. A prior issue of planning is to deal with time constraints, resources constraints, dynamic requests, and adaptability. All these aspects can deal only with a very robust model able to both cope and frame up prior knowledge with probability and other analytic models. The outcome of the planner will be a exible temporal plan of parallel activities distributed within a long horizon. For instance, a possible plan could be: start moving to site X between time 0 and 5, end moving to site X between 10 and 1000, start acquiring frames between 20 and 2000, end acquiring frames between 100 and 4000; start monitoring for interesting locations between 5 and 10, etc. 5.2.2 Scheduling and reactive planning (SRP)

The second component is the scheduling and reactive planning. Reactive planning has to primarily cope the management of resources and requests from all the medium and high level system requests. The SRT is to provide a model-based executive layer interfacing the MP functionalities with the low-level system. It integrates the Mission Planner with the Executive System and keeps the adherence (and synchronization) of the mission level state w.r.t. the execution context. The SRP system will reason over resource bounds within the generated exible temporal plan of the science activities, therefore it has to manage updates, re-planning and the possible inconsistencies that can come out after updating and reconguring the current state. In order to keep the synchronization between the SRP and the reactive state, an internal clock is to update the domain temporal constraints for each tick. For instance, given the plan in the example above, if the rover cannot start moving within time 5, the time constraints are violated and the scheduler is to react in order to recover. Concurrent temporal programming languages will be deployed to specify the behavior of the SRP module. These executive programming languages are endowed with declarative constructs capable to access the temporal model of the system activities. The SRP activity can be summarized as follows (see Planning Schema in Figure 2): Input. Executive state agent: time and resources cycles. Input. Temporal model of the executive components (temporal and resources constraints). Input. Partial mission plan. Activity. Monitoring tasks, checking model consistency, and resources optimization. Output. Cycle of scheduling management The schema of the rst two planning components is shown in the gure 2. Both the MP and SRP shear the time and cause-eect constraints and a priory knowledge, the MP generate a exible temporized plan which is to be managed by the SRP component. High level programs, libraries of precompiled plans, and learned strategies are deployed in order to keep the consistency of the activities scheduled on parallel timelines. 5.2.3 Learning
Both the planning components, namely the Mission Planning (MP) and the scheduling and reactive planning (SRP) are endowed with a sound interface to the other system components through two main learning activities: Learning dynamic behaviors and learning saliency maps.
Figure 2: Planning Schema 5.2.4 Learning dynamic behaviors (LB)
LB is to learn parameters of the current state- behaviors and predict the robot internal states. In this case learning techniques are deployed in order to adapt the declarative models to the real behavior of the represented system. Given a timeline of possibly concurrent activities (e.g. taking measurements, enhance map quality, grab frames, move to an interesting and promising location, exploration etc), and resources requests ( hazard/obstacle avoidance, navigation cameras, telemeter, pan-tilt unit, local and global map, etc.), each one specied by its load and weight as functions of time (initial time and duration) and expected memory/CPU(cycles) consumption, the problem is to optimally allocate the current requested resources so as to maximally satisfy each scheduled activity. The monitor holds a window Wt on the timeline, inspecting the current activities and those needing to take place in the next time interval, and we assume that each activity is labelled according to the time interval in which it cannot be stopped nor delayed. For each activity, in principle, a set of possible resources are assigned, under critical conditions, e.g. needs to use cooperatively both telemeter and Cam while trekking. However it is possible to classify, by minimizing the load and weights distances resources distribution, therefore given an initial partition t0 , of resources, and given that for each activity (Ai ) is suitably initialized, we can obtain the probability of the assignment for each resources to a specic activity by reiterating the computation of and , given that p(rk (tw , l(wk ))) Aj ) = 1 ( l(wk ) ) e , = Zw p(rk (tw , l(wk )))(l(wk ) )2

Where ZW is a normalization factor depending on the window Wt. Clearly this might imply that activities, given the constraints, can be rearranged, on the time line. The LB activity is (Schema LB & LSM in gure 3): Input. 1. Timelines , time constraints, scheduling map. Input. 2. High Level Perception System. Memory 1. Basic behaviors Memory 2. Associative memory of risks, natural actions and exceptions. 6
Figure 3: Schema LB LSM Activity 1. Produce an hypothesis on the best sequence of states given the observed behaviors, anticipate possible states to inform the scheduling activity (in the mission planning task) Output. Resources and time constraints.
Learning saliency map (LSM)
LSM is to identifying interesting object in the current domain and producing a best hypothesis for the current state. The LSM activity is (Schema LB and LSM in gure 3): Input. 1. local and global map, Cam. Input. 2. High Level Perception System. Memory 1. DB of victims features and associated conditional probabilities Memory 2. Associative memory of sites and features. Activity 1. Produce a hypothesis on interesting objects in the surroundings. Output. Qualitative saliency map.
Mission Level Architecture
The functionalities described above are integrated in the architecture illustrated in the gure 4. The Mission Planning Component is connected to the reactive system by means of the Scheduling and Reactive Planning Component. In particular, MP and SRP shear the temporal representation of the executive state which is continuously updated by the rover activities (events). The constraint manager is to check the consistency between the partial plan (current plan) generated by the MP and the tasks scheduled by the SRP. If the current partial plan becomes invalid because of some not nominal behavior, a replanning activity 7

Figure 4: Schema LB LSM

is needed. Instead, if the current plan is robust enough, the MP can be deployed to extend it considering some of the pending goals. The SRP is to mediate between the low-level reactive behavior (events) and the goal-oriented behavior (mission and tasks requests) taking into account the resource (and temporal) constraints.

Localization and Mapping

Low level software architecture in Doro includes also the Localization and Mapping Process (in the rest of the document well name it SLAM). In Doro we realized an intelligent SLAM system: it synthesize a series of tasks useful for high level modules (well explain better later).

Sensor Fusion and Local Metric Mapping
The rst step in a Mapping cycle is Sensor Interpretation. Sensor data are mapped onto local occupancy value. DORO builds a Local Metric Map (robot centered) using only current sensor readings. In this basic step, sensors more frequently used are the Range Finder family (as Sonar, Laser or Infra-red). DORO is equipped with 8 Sonar ranging 180 in forward direction. The occupancy value of each local cell is a function of the distance from cell center and the Sensor Polygon. Sensor Polygon is built with all eight sensors data in this way: each sensor value becomes a point outside the robot, the points are connected with seven solid Lines (connecting each pair of nearest points), rst and last sonars are connected with the center of DORO with a dashed line. Occupancy value of each local cell (the center of) is a costumed function of both distance and relative position with the polygon (solid and dashed line). Sonar readings are used in a cyclical way because these sensors are in continuous polling. In subsection 6.5 well show how SLAM system chooses when (and where) take (and discard) the Local Metric Map made by sonars. Not only sonars are used to build a Local Metric Map: Doro is equipped by a DISTOTM Telemeter on the head that can be used under request(not in continuous polling) as a scanner. The algorithm that builds a Local Map is the same and it takes only an array of point outside the robot and builds the Local Map. This is another Intelligent Task performed by the SLAM System: when (and where) to ask for an absolute control of the head (Pan/Tilt and Telemeter). Local Metric Maps (and Global too) is realized by a C++ Class named OccupancyGridT.
Bayesian Filtering and Global Metric Mapping
Local occupancy value must be integrated in time to build a single, global, Metric Map using a Bayesian Filter as explained in [Thr98]and [Thr02]. A Bayesian Filter is a recursive estimator we use to calculate sequences of posterior probability distribution over a quantity that cannot be observed directly: the Map. We assume that the environment is static, so the true Map can be assumed as a constant (but not our knowledge about it). We name bt the subjective belief about the state of the cell (x,y), i.e. the probability x,y that the cell is occupied conditioned of sensor readings: bt = P (occx,y /ot , st ) x,y

Where o is the observation sequence and s is the robot state (x,y,) sequence (Robot Path). This desired probability can be computed in the following way: bt = + x,y P (occx,y ) 1 P (occx,y ) P (occx,y /oi , si ) 1 P (occx,y ) 1 P (occx,y /oi , si ) P (occx,y ) i=1
Where oi and si represent observation and robot position at ith time step. The last equation can be written in term of b/(1 b), that leads to:

Localization Problem

bt P (occx,y ) x,y = 1 bt 1 P (occx,y ) x,y This is a recursive formula:
P (occx,y /oi , si ) 1 P (occx,y ) 1 P (occx,y /oi , si ) P (occx,y ) i=1
bt1 bt P (occx,y /oi , si ) 1 P (occx,y ) x,y x,y = t 1 bx,y 1 btP (occx,y /oi , si ) P (occx,y ) x,y
Where the last term is the prior probability: if set to 0.5 for each cell, it can be omitted. The left hand side of the last equation ranges from zero to innity, when the belief ranges from zero to one. We use this term in the Global Map. Global Metric Map is realized by the same C++ Class of every kind of Local Metric Map.
Doro is equipped with a pair of Encoder (one for each actuated wheel) and an Inertial Platform. Our goal is to integrate the estimated position of each sensors and to take care of what Doro see while exploring. In rst approximation we can assert that our localization problem is to correct odometric errors (due to slippage and drift) and Inertial Platform errors (due to vibration and electromagnetic elds): this kind of problem is often named, in literature, Position Tracking Problem where errors are assumed small. In contrast with Global Position Problem when a Robot has a map of its environment and it is able to localize everywhere on it. this isnt our problem (at the beginning). We do not solve a Global Localization Problem because DORO is designed for working with no knowledge of the environment: theres no a priori Map in his initial knowledge. Our approach follows the idea in [Thr98] and it is not an approach purely probabilistic: at each computation step, the perceptual error is calculated (estimated), and then corrected. No error distribution will be maintained in memory. This works well when odometers and inertial platform errors are not so big (as in our case). These are restrictive hypotheses because a precise position is needed to build a consistent map. This estimation process tries to minimize the following functional: J

2 o 2 = +1 [(xo robot xrobot ) + (yrobot yrobot ) ] o +2 (robot robot )2 pi +3 [(xpi xrobot )2 + (yrobot yrobot )2 ] robot pi +4 (robot robot )Corr(xrobot , yrobot , robot ) 6 ((wall , robot , new ))
Where the symbols 1 ,. , 6 represents positive parameter to be tuned for minimization. The rst two terms measures the odometric error, the second two measures the inertial platform error and the last two is respectively a measure of the correlation between the local metric map and the global one (a measure of map matching), and a measure of the alignment between robot and walls. In this way DORO can correct small errors and it works well under the constraint that localization processes will be done frequently (that implicitly means small error each time). Minimization of functional J is done with the well known Gradient Descent algorithm. This is what Dodo does today. We are working to make this Localization Process more intelligent as explain in subsection Intelligent Tasks. Now we are testing sensor goodness, in translational and rotational errors, to optimizing beta factors.

Intelligent Tasks

Now we want to show some of our Intelligent tasks accomplished (or to be accomplished) by DORO SLAM System. Our goal is that SLAM System should supply to the other systems detailed information about 10


DORO state: for example nearest unexplored location on the map (that can be used by a planner as default behavior). This task is accomplished by an independent thread under the SLAM Process named Value Iteration Thread as the relative algorithm. Another intelligent task is the integration of various nature Local Map into the Global One: things needed are only same interface (i.e. Class OccupancyGridT). Our actual implementation is based on Sonars (Local Mapping using a circle readings around DORO), Telemeter single reading (Local Mapping using only one reading in one precise direction) and Telemeter circle readings (Local Mapping using telemeter as a laser scanner). Obviously the same Local Mapping algorithm could be implemented with a laser range nder or a radar scanner but DORO hasnt so much exteroceptive sensors. We have told about Telemeter used as a scanner or single reading. First way to use must be delivered by an intelligent decision from SLAM System because Head structure (Pan/Tilt, Telemeter, Cams) is a shared resource. So SLAM System listen continuously request status of the Head and when it looks to be free of order and useful for Mapping, Head System can be Locked by SLAM System to perform a scan Telemeter readings. This must be done only when theres no needs, from Vision System, to acquire Cam information and when DORO has no needs to move. Second way to use Telemeter is passive reading of each sensorial acquisition: while DORO moves inside its environment, each sensors is in continuous polling. Each Telemeter reading passes through a decisional process (who is part of SLAM System) that determines if the current reading should be used for mapping or not. This leads to an important consideration: DORO needs something that can decide if a situation (DORO status and data acquired) is interesting for Mapping or Not. SLAM System Implement a Control Routine that evaluates, for each situation, what should be done with read data. In unknown environment exploration, everything you see is new information and must be unconditioned added to your knowledge, while in known environment you must match what you see and what you should see to determine if theres something wrong. DORO SLAM System implements a routine to determine if it know enough about local environment and if it can use read data to conrm its knowledge without adding new knowledge. This is done to avoid error caused from a Non-Markovian environment where errors does not compensate each others in means (where measurement errors are not independent) but not only: it realizes an human way of reasoning based on how I am sure of what I know! Technically speaking, SLAM Control Subsystem analyzes DORO path (sequence of mapping positions during its Life) and actual position to decide if data read must be used to increase current knowledge or to conrm current localization. Each decisional process inside SLAM System can be supervised by high-level Doro System (or by an operator, if DORO is acting in supervised mode) that can enable or disable intelligent component of SLAM System.

Reactive Navigation

The reactive navigation is based on a behaviour based Fuzzy Logic Controller[Zad65] (FLC). Reactive aptitudes are intended to shape high-level commands in real-time, thus leading to a virtual model of the world; further these aptitudes are built in low-level motor advice, forged on the real world perception. Behaviours[Ste94, Bro90, Rus90] are modelled via a set of fuzzy rules. Dierently from classical control methods like articial potential eld, edge detection and PID controllers, the FLC guarantees robust navigation in a dynamic and unstructured real world, such as indoor environments, and it naturally copes with data uncertainty and does not require an accurate model of the environment. A behaviour based navigation system can be characterized by few but signicant elements, these are: 1. The set of behaviours that the system is able to perform; 2. How concurrent behaviours are merged together in order to produce a single motion command.


In DORO navigation system we distinguish between published and internal behaviours to outline the non one-to-one mapping between the set of fuzzy rule bases implemented and the high level behaviour exposed to the external Mission Planner level. Rule bases are kept simple and light allowing a fast defuzzycation process and an easy and modular composing of new higher level behaviours from simpler ones. The complexity of higher level behaviour (like the ones that we are going to list in the following) is achieved by a smart and accurate sequencing of this elementary fuzzy building blocks into a logical structure that we call behavioural stack. This logical structure helps to keep note of the current low level fuzzy behaviour and the whole state of the navigation (i.e. Resuming a wall-following task once an obstacle has been avoided.). In everyday situations the FLC instinctively shuns obstacles while the robot achieves actions proposed by the cognitive level; there are dierent behaviours that can be performed by the FLC, the most important are: 1. Obstacle avoidance; 2. Point to point; 3. Path tracking; 4. Wall following. 5. Wandering. Obstacle-avoidance behaviour is intrinsically endowed in every action; for each high level command the fuzzy inference system (FIS) includes a specic rule-base that is able to deal with the compromise between avoiding impacts and reaching the given goal. A collision free navigation is always guaranteed. In the FLC we can discriminate essentially between the obstacle-avoidance activity and all the other behaviours that can be viewed like particular cases, with appropriate tunings and data pre-elaborations, of a point-to-point navigation. Furthermore, an intelligent navigation system should be able to embody more than a simple reactive behaviour. To accomplish this, an inference engine must be able to extend his knowledge beyond a strict local and instantaneous perception domain, that prevents any intelligent behavior. Basically,DORO navigation system is built upon two main layers: a Local Perceptual System and a semi-local Navigation System. While the rst reacts in a reexive fashion to events perceived by sensors, the latter dialogs with the mapping module (which owns the local and global map). The collected metrical map, may cover or not the information that the navigation system needs in order to nd a clear path. If such information is available (the robot has passed at least once in that region), the system has more chances to guess which is the most promising direction to take.


[Bro90] Rodney A. Brooks. Elephants dont play chess. Robotics and Autonomous Systems, 6(1&2):315, June 1990. [GV02] Johnson Gamma, Helm and Vlissides. Design patterns: abstraction and reuse of object-oriented design. 2002. [Rus90] E. H. Ruspini. Fuzzy logic in the Flakey robot. In Proc. of the Int. Conf. on Fuzzy Logic and Neural Networks (IIZUKA), pages 767770, Iizuka, JP, 1990. [Ste94] L. Steels. The articial life roots of articial intelligence. Articial Life, 1:75110, 1994. [Thr98] S. Thrun. Learning metric-topological maps for indoor mobile robot navigation. Articial Intelligence, 99(1):2171, 1998.
[Thr02] S. Thrun. Robotic mapping: A survey. In G. Lakemeyer and B. Nebel, editors, Exploring Articial Intelligence in the New Millenium. Morgan Kaufmann, 2002. to appear. [Zad65] L. Zadeh. Fuzzy sets. Informationa and Control, 8:338353, 1965.



HDV178 VLF510 VX9000 RDV-1 Afipr Signature Game JR Sunfire 1999 1098 S SX300 170S6FB Beach C51 ML-3470D Scala Angeles Mp71 Fifa 06 Ericsson T300 Sk5 Of WAR HTS3610 OT-807 Audioline 25 DM-0530 16800 Z5915 TH6220D Optoma HD70 LX-400 Ensemble Coolpix S520 HVR-Z5E WRX STI CU-CE9GKE SGH-T201G Boss DD-6 MC2660E-A PS50A410c1D Ge MP-110 Olufsen BEO5 EX-Z60 1 0 Aspire 4530 Expresso Plus Gmas6 Utility EDC5310 CX7525 LE22C450 VPL-FX52 SRW208P RB-1092 HC600MLS AR-235 275 Sanitarium Panasonic P50 Electronic TC-TX101 Mitsubishi Colt TX32LE8PA LQ-860 EL-531WH Mustang Bass Review RX-V540RDS HI-545 Singer 7568 V-ROD SP-R7000 HD321KJ-jp1 Dvdr3430V 31 HK3600 LAC-M1500R RL34scps Digitech GNX4 SC150 AVR-3808 A KDC-7090R ON TAP FW-M55 PEM-500 Braun 380 DX8450 DGX-230 Pjtx10E Plantronics CT11 NV-M10B DTR4610-08 Canoscan-4400F AW-HE50HN L72805 KD-G201 Bridge 311 TLP-X100 Nikon N90 LP-XG24 B5722 MS09AH DP-508 Racing Audi A8


manuel d'instructions, Guide de l'utilisateur | Manual de instrucciones, Instrucciones de uso | Bedienungsanleitung, Bedienungsanleitung | Manual de Instruções, guia do usuário | инструкция | návod na použitie, Užívateľská príručka, návod k použití | bruksanvisningen | instrukcja, podręcznik użytkownika | kullanım kılavuzu, Kullanım | kézikönyv, használati útmutató | manuale di istruzioni, istruzioni d'uso | handleiding, gebruikershandleiding



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101