The PEIS Home is an experimental environment in which we can test the concept of PEIS Ecology. The PEIS Home looks like a typical bachelor apartment of about 25 square meters. It consists of a living room, a bedroom, and a small kitchen. The walls have been constructed to be 1.40 meter high, so that the observers can get a bird's eye view of the entire apartment. Experimenters and visitors can observe the behavior of the PEIS Home and of its inhabitants from an "observation deck". The PEIS Home is equipped with a communication and computational infrastructure, and with various sensors and actuators.
Several PEIS live in our PEIS Home, and they can be connected in different PEIS Ecology configurations. The current PEIS include: three mobile robots, a smart refrigerator, an autonomous vacuum cleaner, and a monitoring system using web-cameras mounted on the ceiling. The web-cameras cover the living room and the kitchen, but not the bedroom, which is considered a private space. More PEIS are being continuously added to our PEIS Home.
All PEIS in a PEIS Ecology use a uniform communication and cooperation model. This model has been implemented in a run-time portable library called the PEIS Kernel, which acts as a middle-ware for all components (processes) in the ecology. The PEIS Kernel implements an ad-hoc P2P network using any available communication links between participating PEIS. On the top of this, the PEIS Kernel maintains a distributed tuple-space, together with an event mechanism. The PEIS Kernel is compliant with the dynamic nature of a PEIS Ecology, and it has no dependencies on any centralized point needed to bootstrap the environment.
The PEIS Middleware will be distributed as open source. Version 0.4.0 is currently available to registered alpha-testers: see the Dowloads section of this site.
Progress on the PEIS Ecology project is marked by a series of milestone demonstrations. The following examples, in chronological order, illustrate increasingly complex aspects of the PEIS Ecology.
1. Wake-up Johanna
This simple experiment illustrates the basic communication and cooperation mechanisms in a PEIS Ecology. Originally presented at the sOc-EUSAI conference on Smart Objects and Ambient Intelligence (Grenoble, FR, October 2005).
Scenario: The robot Emil has the task to wake up Johanna. She decides to delegate the execution of this task to the robot Pippi, and generates a plan for Pippi consisting of the three actions: {GoTo(bed), Talk(wakeup), GoTo(home-position)}. Emil sends the first action to Pippi, who executes it. In order to better keep track of her own position in the environment, Pippi uses the information provided by the ceiling camera system. If Pippi exits the field of view of the cameras (eg, in the bedroom) she reverts to using its own (unreliable) odometry for self-localization. When Pippi reaches the bed, she notifies this event to Emil, who then sends her the next action, and so on. When Pippi finally reaches her home position, she notifies Emil and the task is completed.
Read the paper (2.0 MB) Watch the movie (2.6 MB, WMV)
2. The milk has gone bad
This experiment shows that a PEIS Ecology can solve a problem (pervasive olfaction) that would be hard to solve by a mobile robot alone, or by a sensor network alone. Presented at the IEEE Int Conf on Robotics and Automation (ICRA-06, Orlando, Florida, May 2006).
Scenario: At the beginning of the day, the gas sensors in the refrigerator detect a bad smell. A global "home security monitor" decides to execute an investigation plan. It sends to the robot Pippi the order to go to the fridge, and use its sophisticated electronic nose to classify the type of odor. Pippi uses the information from the ceiling camera to self-localize during navigation. When approaching the fridge, Pippi sends it the request to open its door. Pippi then analyzes the odor, by exploiting the context information provided by the fridge itself, who uses an RFID tag reader to access the information about the goods placed inside it. Pippi determines that the milk has gone bad, and it publish this result to the tuple-space. The robot Emil moves to the bedroom and warns Abdel of the situation.
Read the paper (347 KB) Watch the movie (3.2 MB, WMV)
3. Push that box out of the way
This experiment extends the first one above, by introducing a new event: the robot encouters a box blocking the way, and asks it if it can be pushed. The experiment shows a first solution to the anchoring problem: how can the robot determine that the box that it sees in front of it is the PEIS with ID=22? Presented at the IEEE Int Conf on Robotics and Automation (ICRA-06, Orlando, Florida, May 2006).
Scenario: The robot Emil receives the request to take a photo of Malin. She delegates the execution of this task to the robot Pippi, and generates a plan for it consisting of the three actions: {GoTo(bed), Take(photo), GoTo(home-position)}. Emil sends the first action to Pippi, who executes it. When approaching the bedroom, Pippi detects a box blocking the door. She needs to establish if the box can be pushed or not, but she cannot extract this information from her sensors. She therefore tries to see if the box is a PEIS, in which case she could ask this information directly to the it. She compares the observed properties of the box to the "Appearance" property commnicated by each PEIS in the ecology, and she finds that the box is the PEIS with ID=22. She then asks the value of the "pushable" property to this PEIS. The box is pushable, so Pippi pushes it out of the way and completes her task.
Read the paper (2 MB) Watch the movie (3.9 MB, RealVideo)
4. Big robots and tiny devices cooperate
This experiment shows that the PEIS Ecology concept and middleware allow the smooth integration of robotic devices at very different scales. In this experiment, a standard mobile robot with an onboard PC cooperates with small sensor-network motes and with micro-controller boards. Presented at the IEEE/RSJ Int Conf on Intelligent Robots and Systems (IROS-07, San Diego, California, October 2007).
Scenario: A Mote installed on a plant near the window detects that the plant is receiving too much direct sunlight, and sends a shut request to the window blinds, which are equipped with a motor and a microcontroller. Soon after, a mobile robot called Astrid is given the task to take a photo of the TV table. When Astrid reaches the table, it realizes that the light is insufficient. Since the window blinds cannot be open due to the conflicting request from the plant, Astrid sends a switch-on request to lamp on the table, which itself contains a Mote. Astrid can now take the photo and complete the task.
Read the paper (294 KB) Watch the movie (3 MB, RealVideo)
5. A PEIS Ecology in Korea
This experiment demonstrates the porting of the PEIS Ecology concept and middleware to a different physical platform. The platform is the demostration home of ETRI in Dejon, South Korea, one of the partners in the PEIS Ecology Project.
Scenario: The scenario is a combination of the above examples 3 and 4. A Mote (PEIS-1) installed on a plant near the window detects the amount of light received by the plant, and sends open or shut requests to the window curtains, which are equipped with a motor connected to a microcontroller (PEIS-2). A mobile robot called Weaver (PEIS-3) is given the task to wake up Joo-Hanna: it sends a turn-on request to the lights in the bedroom, controlled via a Zigbee interface, and then navigates to the bedroom. During navigation, Weaver detects an obstacle blocking the door. Weaver retrieves the status of the door via a Zigbee interface, and finds that the door is open. It therefore pushes the obstacle away, wakes up Joo-Hanna, and completes the task.
Read the paper (294 KB) Watch the movie (4.9 MB, Mpeg4)
6. Dynamic self-configuration and re-configuration
This experiment shows the ability of a PEIS Ecology to self-configure, at execution time, to perform a given task using the currently available resources; and to automaticaly re-configure to cope with errors and changes in the environment. Presented at the IEEE/RSJ Int Conf on Intelligent Robots and Systems (IROS-07, San Diego, California, October 2007).
Scenario: When the morning paper arrives, Astrid the robot is assigned the task to wake up and deliver the newspaper to Jay. Astrid, generates the plan to: move to entrance, get newspaper, move to bedroom, give wakeup-signal. Astrid acquires the state of the ecology, and finds a configuration to execute the first action using the cameras in the ceiling to track its own position. Near the entrance, however, Astrid exits the field of view of the cameras, and the ceiling tracker signals a failure. Astrid acquires the new state of the ecology, and decides to use a configuration in which its own laser (more energy expensive than the cameras) is used for localization. Astrid is then given the newspaper, it navigates to the bedroom, wakes up Jay and gives him the newspaper.
Read the paper (411 KB) Watch the movie (3.4 MB, RealVideo)
7. Final demo of the ETRI project
This experiment summarizes the scientific achievements of the PEIS Ecology project until 2007. It illustrates in particular: the ability to integrate highly heterogeneous devices (a robot, a lamp controller, and tagged book); the ability to effectively share information among these PEIS; the ability to automatically configure the ecology to perform a given task; and the ability to dynamically re-configure in case of failures. Presented at the IEEE/RSJ Int Conf on Intelligent Robots and Systems (IROS-08, Nice, France, Sep 2008).
Scenario: Alex lays in bed with a broken leg, and he asks the PEIS Ecology to bring him his favorite book. The request is dispatched to the robot Astrid. Astrid acquires the current state of the Ecology from the tuple-space, and it finds that the book has been detected by the RFID tag reader on the kitchen table. Astrid then generates a plan to: turn on lights, go to kitchen, dock to table, take book, go to bedroom, give book. For each action, Astrid configures the Ecology in a suitable way and starts it. While performing the go to bedroom action, the ceiling tracking system fails: so Astrid generates an alternative configuration for that action, using its own laser for localization instead of the ceiling system. Astrid finally reaches the bed and delivers the book to Alex.
Read the paper (1.6 MB) Watch the movie: RealVideo (3.5 MB) or MP4 (5 MB)
8. Ecological delivery of a drink
This experiment was the result of the Integrated Project Work of the second year students of the International Master's in Robotics and Intelligent Systems, in 2008. The students were Li Feng, Nikolay Kramarev, Wenjie Lin, Wang Min, Andreas Persson, Tom Ranskär and Jonas Ullberg. The experiment is an ecological realization of the usual "robot butler" task of serving a drink to a guest.
Scenario: A person enters the PEIS-Home and sits on the sofa. A stereo tracking camera mounted on the ceiling recognizes the presence of the person and supplies a coarse estimate of his position. A mobile robot is then dispatched to the person's approximate position to identify the guest with the help of an on-board pan-tilt camera and a face recognition algorithm. In this scenario it is assumed that the system knows the favorite drink of a set of frequent guests. The PEIS-Table is sent towards the fridge to fetch the guest's favorite drink. The PEIS-Fridge, equipped with an internal gripper, an internal camera and an actuated door places the drink on the table, which has in the meanwhile docked the open fridge. The table then navigates towards the person's current position to deliver the drink.
Read the paper (1.6 MB) Watch the movie: Flash Video (20 MB)