How can Robots learn from Honeybees? Karl Crailsheim, Ronald Thenius, ChristophMöslinger, Thomas Schmickl Apimondia 2009, Montpellier Beyond robotics
Definition of robot : Robots A device that automatically performs complicated and often repetitive tasks(merriam-webster) Any automatically operated machine that replaces human effort(encyclopaedia Britannica)
Anew approach to the coordination of multi-robot systems Swarms with large numbers of mostly simple physical robots Collective behavior emerges from: interactions between the robots interactions of robots with environment Swarm intelligence Swarmrobotics
Application Two behaviors of bees applicable for bioinspired Swarm Robotics: 1.Optimum Finding (Collision-based Algorithm) Aggregating at bright or warm spots inside arena Tested in simulations and real robots 2.Navigiation(Trophallaxis-based Algorithm) Finding optimal paths from nest to food Tested in simulations, ready for real robots
1. Optimum Finding(Collision-based) Behaviour of bee(s) in a temperature arena (120 seconds with 10 second snapshots) Single bee 30 bees Warm spot Warm spot
1. Optimum Finding(Collision-based) Possible algorithm: Analyzingthebehaviour Robots move randomly Robotstopswhen encountering another robot When stopped, the robot measures the illuminance Thehighertheillum.the longer the robot waits
1. Optimum Finding(Collision-based) Possible algorithm: Bee walks randomly Analyzingthebehaviour Beestopswhen encountering another bee Whenstopped, thebee measures the temperature Thehigherthetemp.the longer the bee waits
1. Optimum Finding(Collision-based)
1. Optimum Finding(Collision-based) Conclusion Collision-based strategy is: suitable for aggregation scenarios robust malfunctioning robots have little influence robot sensors can be imprecise able to differentiate different light spots adaptive in dynamic scenarios (re-decision)
Relevant Posters Robustness of group decisions in honeybees SibylleHahshold, Gerald Radspieler, Ronald Thenius, Thomas Schmickl& Karl Crailsheim The influence of group size on cooperative decision making in honeybees Martina Szopek, Gerald Radspieler, Ronald Thenius, Thomas Schmickl& Karl Crailsheim
2. Navigation (Trophallaxis-based) Purpose of trophallaxis in honeybees: Food storage(and processing) Feeding (queen, hivemates, drones) Transfer of information
2. Navigation (Trophallaxis-based) Trophallaxis with feeder comb gradient lousy weather! comb comb bee hive
2. Navigation (Trophallaxis-based) Why Trophallaxis Strategy? To create a map of the environment cooperatively. Swarm robots: are small have limited range for sensors and communication have simple and inaccurate sensors are numerous Trophallaxis strategy: requires only little computational resources requires only neighbour-toneighbour communication works even with imprecise information the more robots, the better
2. Navigation (Trophallaxis-based) 1. Addition-rate (+a) Trophallaxis in a robotic swarm 2. Transfer-rate (±t) 3. Consumption-rate (-c) Results in a distributed gradient map Robots can move uphill or downhill this gradient
2. Navigation (Trophallaxis-based) Collective perception in a robot swarm Target 2 (small) Target 1 (big) red robots = low virtual nectar; white robots = high virtual nectar Trophallaxis strategy enables the swarm to distinguish between different target sizes! Thomas Schmickl, Christoph Möslinger and Karl Crailsheim Second SAB 2006 International Workshop on Swarm Robotics, Rome, Italy.
2. Navigation (Trophallaxis-based) Bio-inspired Navigation of Autonomous Robots in Heterogenous Environments dump unsuitable terrain Loaded robots are able to circumvent unsuitable terrain because it is visible in the distributed map Thomas Schmickl, Christoph Möslinger, Ronald Thenius and Karl Crailsheim Emerging Technologies, Robotics and Control Systems Vol. 2, International Society for Advanced Research.
2. Navigation (Trophallaxis-based) Individual adaptation allows collective path-finding in a robotic swarm dirt dump dirt unsuitable terrain Swarm is able to differentiate the orientation of the unsuitable terrain; loaded robots take the fastest path Thomas Schmickl, Christoph Möslinger, Ronald Thenius and Karl Crailsheim Emerging Technologies, Robotics and Control Systems Vol. 3, International Society for Advanced Research.
2. Navigation (Trophallaxis-based) Conclusion The whole is greater than the sum of its parts Trophallaxis strategy enables swarm to: generate a distributed gradient map perform aggregation scenarios differentiate target sizes perform cleaning scenarios avoid unsuitable terrain distinguish between different kinds of unsuitable terrain choose optimal paths in complex environments