File preview
Multi-Robot Search and Rescue:
An Open-Ended Bridge Between Theory and Practice
Philip Twu and Magnus Egerstedt
Georgia Institute of Technology (philip.twu@jhuapl.edu, magnus@gatech.edu)
http://gritslab.gatech.edu/home/
Purpose
Teach students how to methodically apply networked control theory concepts learned in the classroom towards solving complex cyber-physical system (CPS) engineering problems.
Virtual Environment
1. Models Coupling of Controller With Physical Limitations • Actuator saturation • Finite sensing radius • Finite sensor resolution • Finite update rates 2. Models Interaction Robot With Physical Environment • Dynamic network topology • Robot-obstacle collisions • Robot-robot collisions • Sensor noise 3. Developed for MATLAB Programming Environment
Educational Outcomes
1. Students Learned to Extend Computationally-Focused Lecture Material to Handle CPS Engineering Challenges • “How to actually design for real situations.” • “Implementation is much harder than what I believe it would be based just on the material from lecture.” • “Theory and practice differ quite a bit, the encoding of weights in MATLAB had to be done in a different form than in theory.” 2. Reinforced Key Concepts Learned Throughout Semester • “Combine different concepts in class for a working system.” • “Understand the ideas/subjects learned in class more rigidly.” 3. Fostered Need to Continue Learning Outside of Class • “There’s a lot of ways to solve each problem, more research needs to be done to find optimal solutions.” • “I also had to research and learn into concepts that weren’t exhaustively covered in class.”
Approach
1. Design a Computational Solution • Students derive high-level decentralized control laws to coordinate multiple robots for a search and rescue mission. 2. Evaluate Against Physical System and Environment • Control laws are scripted and tested in a custom-designed virtual environment which models physical domain effects. 3. Iteratively Engineer a Cyber-Physical Solution • By toggling physical domain effects on one at a time, students adapt theoretical solutions based on simplified mathematical models to obtain solutions to CPS problems.
Mission Challenges
Project Instructions
The year is 2030 and NASA has identified an asteroid that is on a collision course with Earth! Robot scouts sent previously to collect surface samples have been temporarily disabled from a pulse of EM radiation. Based on your extensive experience in networked controls, members of the robotics faculty have selected you to lead a rescue mission. Your mission is to design decentralized controllers to: 1. Navigate a team of 6 robots through the asteroid’s terrain 2. Locate & reactivate 6 disabled robots from a previous mission 3. Bring all 12 robots back to the platform, get into formation, and await to be picked up by a nearby orbiting spacecraft. Good luck, the future of humanity rests in your hands!
1. Dynamic network topologies
2. Robot collision avoidance
3. Topology switching
4. Increased Student Engagement and Excitement • Videos of top 5 fastest solutions are shown in last lecture as students cheer and applaud for their fellow classmates. • 100% of students responding to survey suggested the project be included in future semesters the course is taught.
5. Successful Integration into Existing Course • Deployed in Fall 2010 and Fall 2011 semesters of ECE8823 Networked Controls at the Georgia Institute of Technology.
4. Obstacle avoidance
5. Searching and coverage
6. Decentralized coordination
Distribution
Detailed instructions for assigning the multi-robot search and rescue project, along with MATLAB source code for the virtual environment, are publicly available for download for anyone interested in adopting a similar project within their own classes:
7. Disconnected networks 8. Splitting and merging 9. Formation assignment
http://users.ece.gatech.edu/~magnus/ece8823finalproject.html
An Open-Ended Bridge Between Theory and Practice
Philip Twu and Magnus Egerstedt
Georgia Institute of Technology (philip.twu@jhuapl.edu, magnus@gatech.edu)
http://gritslab.gatech.edu/home/
Purpose
Teach students how to methodically apply networked control theory concepts learned in the classroom towards solving complex cyber-physical system (CPS) engineering problems.
Virtual Environment
1. Models Coupling of Controller With Physical Limitations • Actuator saturation • Finite sensing radius • Finite sensor resolution • Finite update rates 2. Models Interaction Robot With Physical Environment • Dynamic network topology • Robot-obstacle collisions • Robot-robot collisions • Sensor noise 3. Developed for MATLAB Programming Environment
Educational Outcomes
1. Students Learned to Extend Computationally-Focused Lecture Material to Handle CPS Engineering Challenges • “How to actually design for real situations.” • “Implementation is much harder than what I believe it would be based just on the material from lecture.” • “Theory and practice differ quite a bit, the encoding of weights in MATLAB had to be done in a different form than in theory.” 2. Reinforced Key Concepts Learned Throughout Semester • “Combine different concepts in class for a working system.” • “Understand the ideas/subjects learned in class more rigidly.” 3. Fostered Need to Continue Learning Outside of Class • “There’s a lot of ways to solve each problem, more research needs to be done to find optimal solutions.” • “I also had to research and learn into concepts that weren’t exhaustively covered in class.”
Approach
1. Design a Computational Solution • Students derive high-level decentralized control laws to coordinate multiple robots for a search and rescue mission. 2. Evaluate Against Physical System and Environment • Control laws are scripted and tested in a custom-designed virtual environment which models physical domain effects. 3. Iteratively Engineer a Cyber-Physical Solution • By toggling physical domain effects on one at a time, students adapt theoretical solutions based on simplified mathematical models to obtain solutions to CPS problems.
Mission Challenges
Project Instructions
The year is 2030 and NASA has identified an asteroid that is on a collision course with Earth! Robot scouts sent previously to collect surface samples have been temporarily disabled from a pulse of EM radiation. Based on your extensive experience in networked controls, members of the robotics faculty have selected you to lead a rescue mission. Your mission is to design decentralized controllers to: 1. Navigate a team of 6 robots through the asteroid’s terrain 2. Locate & reactivate 6 disabled robots from a previous mission 3. Bring all 12 robots back to the platform, get into formation, and await to be picked up by a nearby orbiting spacecraft. Good luck, the future of humanity rests in your hands!
1. Dynamic network topologies
2. Robot collision avoidance
3. Topology switching
4. Increased Student Engagement and Excitement • Videos of top 5 fastest solutions are shown in last lecture as students cheer and applaud for their fellow classmates. • 100% of students responding to survey suggested the project be included in future semesters the course is taught.
5. Successful Integration into Existing Course • Deployed in Fall 2010 and Fall 2011 semesters of ECE8823 Networked Controls at the Georgia Institute of Technology.
4. Obstacle avoidance
5. Searching and coverage
6. Decentralized coordination
Distribution
Detailed instructions for assigning the multi-robot search and rescue project, along with MATLAB source code for the virtual environment, are publicly available for download for anyone interested in adopting a similar project within their own classes:
7. Disconnected networks 8. Splitting and merging 9. Formation assignment
http://users.ece.gatech.edu/~magnus/ece8823finalproject.html