1. Evolution, Self-organization and Swarm Robotics Minsu Kim

2. Contents 1. Introduction 2. Evolutionary Design of Self-organizing Behaviors 2.1 The Design Problem 2.2 Evolution of Self-organizing Behaviors 3. Studies in Evolutionary Swarm Robotics 3.1 A Swarm Robotics Artifact: The Swarm-bot 3.2 Synchronization 3.3 Coordinated Motion 3.4 Hole Avoidance 4. Conclusion

3. 1. Introduction What is the Swarm robotics? -A particular class of multi robot systems. -It emphasizes aspects like decentralization of control, robustness, flexibility and scalability. -It is often inspired by the behaviors of social insects -> these are definitely beneficial for a swarm of autonomous robots. -Many activities carried out by robots are the result of self-organizing processes. -By designing for self-organization, only minimal complexity is required for each individual robot, and the whole system can solve a complex problem in a flexible and robust way.

4. 1. Introduction What is the Swarm robotics’ problem? -Because the relationship between simple local rules and complex global properties is indirect, the definition of the individual behaviors is particularly challenging way. ->The solution proposed in this paper relies on artificial evolution as the main tool for the synthesis of self-organizing behaviors. http://www.swarmanoid.org/index.php https://www.youtube.com/watch?v=CJOubyiITsE

5. 2. Evolutionary Design of Self-organizing Behaviors 2.1 The Design Problem The challenge is given by the necessity to decompose the desired global behavior into simpler individual behaviors.

6. 2. Evolutionary Design of Self-organizing Behaviors 2.1 The Design Problem The decomposition from the global to the individual behaviors could be simplified by taking inspiration from natural systems. It is not always possible to take inspiration from natural processes because they may differ from artificial systems in important parts and there are no natural processes that can be compared to the artificial one.

7. 2. Evolutionary Design of Self-organizing Behaviors 2.2 Evolution of Self-organizing Behaviors Evolutionary robotics represents an alternative approach to the solution of the design problem. By evaluating the robotic system as a whole, it eliminates the arbitrary decompositions at both the level of finding the mechanisms of the self-organizing process and the level of implementing those mechanisms into the rules that regulate the interaction between robot and the environment. ; Evolutionary robotics is a methodology that uses evolutionary computation to develop controllers for autonomous robots.

8. 3. Studies in Evolutionary Swarm Robotics 3.1 A Swarm Robotics Artifact: The Swarm-bot “s-bot”: a unit of the “swarm-bot” Traction system: track & wheels(treel) Turret Gripper: open, close, lift Infrared proximity sensor Proximity sensor: to perceive terrain’s roughness Light sensor, Temperature/humidity sensor A three-axis accelerometer Encoder Omnidirectional camera, Colored LED, Microphone, Speaker: to detect and communicate with other s-bots Torque sensor, traction sensor: to detect the direction and the intensity of the pulling and pushing forces.

9. 3. Studies in Evolutionary Swarm Robotics 3.1 A Swarm Robotics Artifact: The Swarm-bot

10. 3. Studies in Evolutionary Swarm Robotics 3.2 Synchronization In this paper, robots exploit communication in order to synchronization. Robots communicate through sound pulses that directly reset the internal oscillator designed to control the individual switch from homing to foraging and vice versa.

11. 3. Studies in Evolutionary Swarm Robotics 3.2 Synchronization - Experimental Setup We aim at studying the evolution of behavioral and communication strategies for synchronization. Experiment requires that: - Each s-bot displays a simple periodic behavior, that is, moving back and forth from a light bulb positioned in the center of the arena. - s-bots have to synchronize their movements, so that their oscillations are in phase with each other. In order to communicate with each other, s-bot are provided with a signaling system, which can produce a continuous tone with fixed frequency and intensity.

12. 3. Studies in Evolutionary Swarm Robotics 3.2 Synchronization - Experimental Setup The evolutionary algorithm is based on a population of 100 binary- encoded genotypes. The population is evolved for a fixed number of generations, applying a combination of selection with elitism and mutation. At each generation, the 20 best individuals are selected for reproduction and retained in the subsequent generation. Simulation environment

13. 3. Studies in Evolutionary Swarm Robotics 3.2 Synchronization - Results-Scalability of the Evolved Behaviors - The higher the density of robots, the higher the number of interferences that lead to failure.

14. 3. Studies in Evolutionary Swarm Robotics 3.2 Synchronization - Results-Scalability of the Evolved Behaviors -Many controllers perfectly scale, having a performance very close to the mean performance measured with 3 s-bots. -If the perceived signal does not vary in time, it does not bring enough information to be exploited for synchronization. -The lack of locality and of additivity is the main cause of failure for the scalability of the evolved synchronization mechanism. The case to analyze the scalability property only

15. 3. Studies in Evolutionary Swarm Robotics 3.2 Synchronization - Results-Tests with Physical Robots To test the robustness of the evolved controllers, We downloaded an controller onto the physical robots. Synchrony is quickly achieved and maintained throughout the whole trial. With the C13 controller

16. 3. Studies in Evolutionary Swarm Robotics 3.3 Coordinated Motion The coordination ability is essential for an efficient motion of the swarm-bot.

17. 3. Studies in Evolutionary Swarm Robotics 3.3 Coordinated Motion-Experimental Setup At the beginning of a trial, the s-bots start with their chassis oriented in a random direction. s-bots’ goal is to choose a direction of motion on the basis of only the information provided by their traction sensor, and then to move as far as possible from the starting points. We exploit artificial evolution to synthesize a simple feed- forward neural network that encodes the motor commands in response to the traction force perceived by the robot.

18. 3. Studies in Evolutionary Swarm Robotics 3.3 Coordinated Motion-Results-Behavioral Analysis In order to appreciate the behavioral strategy of each individual, the activation of the motor units was measured in correspondence to a traction force whose angle and intensity were systematically varied.

19. 3. Studies in Evolutionary Swarm Robotics 3.3 Coordinated Motion-Results-Behavioral Analysis How a simple feed-forward neural network flow? 1. Each s-bot perceives a single traction force, which roughly indicates the average direction of group’s motion. 2. An s-bot rotates to align to the perceived traction force. 3. Some s-bot will be faster than the other, therefore reinforcing the traction signal in their direction. 4. The other s-bots perceive an even stronger traction force, which speeds up the alignment process. Overall, this positive feedback mechanism makes all s-bots quickly converge toward the same direction of motion.

20. 3. Studies in Evolutionary Swarm Robotics 3.3 Coordinated Motion-Results-Scalability and Generalization with simulated and Physical Robots Tests with real robots showed a good performance as well, confirming the robustness of the evolved controller. Tests are performed in various cases. - four s-bots form linear structure - s-bots are on rough terrain - s-bots are semi-rigid - six s-bots form linear structure - s-bots form a square structure and star shape

21. 3. Studies in Evolutionary Swarm Robotics 3.3 Coordinated Motion-Results-Scalability and Generalization with simulated and Physical Robots

22. Overall, the tests with simulated and physical robots prove that the evolved controllers produce a self-organizing system can be able to achieve and maintain coordination among the individual robots.

23. 3. Studies in Evolutionary Swarm Robotics 3.4 Hole Avoidance

24. 3. Studies in Evolutionary Swarm Robotics 3.4 Hole Avoidance-Experimental Setup For a swarm-bot to perform hole avoidance, two main problems must be solved: - Coordinated motion must be performed in order to obtain coherent movements of the s-bot. - The presence of holes must be communicated to the entire group. We study and compare three approaches to communication. - Direct Interactions setup(DI). - Direct Communication setup(DC). - Evolved Communication setup(EC).

25. 3. Studies in Evolutionary Swarm Robotics 3.4 Hole Avoidance-Experimental Setup We decided to let evolution shape the controller testing the swarm-bot both in environments with and without holes. In all cases, the s-bots start connected in a swarm-bot formation, and the orientation of their chassis is randomly defined.

26. 3. Studies in Evolutionary Swarm Robotics 3.4 Hole Avoidance-Result All evolutionary runs were successful, each achieving a good performance. The evolutionary experiments were replicated 10 times. DI setup: s-bots can rely only on direct interactions, shaped as traction forces. - The s-bots that detect a hole invert the direction of motion, therefore producing a traction force that perceived by the rest of group. - This traction forces are sufficient to trigger hole avoidance. - If the signal of traction force is not strong enough, swarm-bot is not be able to avoid falling.

27. 3. Studies in Evolutionary Swarm Robotics 3.4 Hole Avoidance-Result DC setup: s-bots can rely on both direct interactions shaped as traction forces and direct communication through sound signals. - The s-bots that detect a hole invert their direction of motion and emit a continuous tone. - The s-bots that perceive a sound signal stop moving. - When no s-bots perceive the hole, signaling cease and coordinated motion can start. - In this setup, direct communication reinforces the interactions through traction forces.

28. 3. Studies in Evolutionary Swarm Robotics 3.4 Hole Avoidance-Result EC setup: Similarly to the DC setup, s-bots can exploit both traction and sound signal. - Because evolution is responsible for shaping both the signaling mechanisms and the response to the perceived signals, signaling and reaction strategies to control the speaker are complex. - If a strong traction force is perceived and if a sound signal was previously emitted, signals are not emitted. In both cases, an avoidance action was already initiated.

29. 3. Studies in Evolutionary Swarm Robotics 3.4 Hole Avoidance-Result -The behaviors evolved within the EC setup performs significantly better than those evolved within both the DI and the DC setups. -The behaviors evolved within the DC setup performs better than the DI setup so that we can conclude that the direct communication is clearly beneficial for hole avoidance. - Moreover, evolving the communication protocol leads to a more adapted system.

30. 3. Studies in Evolutionary Swarm Robotics 3.4 Hole Avoidance-Tests with Physical Robots Each selected controller was evaluated in 30 trials. The behavior produced by the evolved controllers tested on the physical s-bots is very good and closely corresponds to that observed in simulation.

31. 4. Conclusion We make sure that ‘Evolutionary robotics’ are suitable methodology for the swarm of autonomous robots.