Biomutualism. Bio-inspiration. Material properties. Robotic Fish. o Passive o Flexibility. o Evolutionary robotics

Transcription

1

2 Biomutualism Bio-inspiration Material properties o Passive o Flexibility Robotic Fish o Evolutionary robotics

3 Design Process Mathematical Models Material Properties Feedback Physics-Based Simulation Optimization Fabrication and Evaluation Rapid Prototyping

4 Study Overview o Optimize Caudal Fin o Dimensions o Flexibility Model Simulation o Physically Validate o Stable velocity o Improve simulation Reality

5 Applications Ecological Monitoring Harbor Surveillance

6 Biological Studies Elicit a response o ex. robot as predator o Predator inspection o ex. robot as leader o Schooling

7 Outline o Introduction o Evolution Park o This Study o Only Flexibility o Flexibility + Dimensions o Conclusion

8 Evolution Park o NSP-Sponsored testbed o Cross department collaboration o Facilities o Robot grab-bag o Compute cluster o 4,500 gallon test tank o Rapid prototyping 3D printer

9 3D Printer Objet Connex350 Prints multiple material

10 Young s Modulus o (Modulus of elasticity) ~ 0.01 GPa o Material property o Higher value! higher stiffness o Lower value! higher flexibility ~ 10 GPa ~ 100 GPa

11 Printed Robotic Fish

12 Printed Robotic Fish o Printed parts o body o gears o fins o Electronics o Arduino o Servo o LiPo battery

13 Outline o Introduction o Evolution Park o This Study o Only Flexibility o Flexibility + Dimensions o Conclusion

14 Study Parameters o Fixed control o 30 amplitude o 0.9 Hz frequency o Flexible, rectangular caudal fin o Swims on the surface

15 Mathematical Model Hydrodynamics Flexibility Net Hydro Force Net Drag Force Instantaneous Hydro Force Wang et. al. 2011, 2012

16 Caudal Fin Example

17 Outline o Introduction o Evolution Park o This Study o Only Flexibility o Flexibility + Dimensions o Conclusion

18 Optimize Only Flexibility Optimization target o Maximal average velocity Hill-climber o 30 runs o 100 candidates tested Evolution o 30 runs o 100 individuals o 100 generations

19 Physical Validation o Stable velocity o Seven trials o Remove best o Remove worst o Compute average

20 Experimental Comparison Model Prediction Simulation Results 3D Printed Materials Improved Model

21 Outline o Introduction o Evolution Park o This Study o Only Flexibility o Flexibility + Dimensions o Conclusion

22 Optimization Dimensions and Flexibility o Evolve o flexibility o fin dimensions BODY Elasticity Height o Maximal average velocity Length

23 Optimization Dimensions and Flexibility o Maximal flexibility for every set of dimensions o Constraints o Length max = 14 cm o Length min = 4 cm o Modulus max = 50 GPa

24 Conclusion o For Evolutionary Computation o Models can approximate flexible materials o Models can approximate hydrodynamics o Multi-material 3D printers can fabricate evolved flexible solutions o EC results can help improve the modeling process o Design process can be repeated for other environments

25 Future Directions o Energy consumption o Morphology o Expand models o Non-rectangular fins o Complex tasks o Speed, maneuverability o Higher level! waypoint following

26 Acknowledgements o Professor Janette Boughman o Dr. Jason Keagy o Dr. Liliana Lettieri National Science Foundation grants CNS , CNS , DBI , CCF , IIS , ECCS , ECCS , CNS U.S. Army Grant W911NF o Members of o o o o The SENS Laboratory The Smart Microsystems Laboratory The DevoLab The BEACON Center

27 thank you

28

29 Big Picture Robotics Underwater Vehicle System Optimization Strategy Control Morphology Propeller Paired Fin Caudal Fin Guess and Check Gradient Climbing Evolutionary Computation Industrial Search and Rescue Sensor Node Us

30 Mathematical Model o Aquatic environment o Reality gap o Model accuracy o Elongated-body theory

31 Outline o Introduction o Evolution Park o This Study o Only Flexibility o Flexibility + Dimensions o Conclusion

32

33 Future Directions Efficiency Power usage Mechanical work Performance Coevolution Control Morphology Complex tasks Multi-Objective