Evolutionary Techniques in Physical Robotics
ICES '00 Proceedings of the Third International Conference on Evolvable Systems: From Biology to Hardware
The GOLEM Project: Evolving Hardware Bodies and Brains
EH '00 Proceedings of the 2nd NASA/DoD workshop on Evolvable Hardware
Evaluation of real-time physics simulation systems
Proceedings of the 5th international conference on Computer graphics and interactive techniques in Australia and Southeast Asia
Crossing the reality gap in evolutionary robotics by promoting transferable controllers
Proceedings of the 12th annual conference on Genetic and evolutionary computation
A fast and elitist multiobjective genetic algorithm: NSGA-II
IEEE Transactions on Evolutionary Computation
| Hi-index | 0.00 |
Transferring designs in evolutionary robotics from simulation to reality remains problematic. It has been addressed by using quasi-static physics simulators, adding noise to encourage robustness, and evolving primarily in simulation then evolving on actual hardware for fine-tuning. This paper experiments with this idea: All physics simulators have errors, but if the errors are distinct, one might profitably use multiple simulators to detect unrealistic physical behavior in simulation. Two physics simulators are used to evolve a controller for quadruped locomotion. Preliminary results validate some assumptions and further work is suggested.