Design team members: Zack Zhu
Supervisor: Professor Mohamed Kamel
Background
Insects have demonstrated remarkable abilities in a number of survival tasks such as foraging and path-finding. Flocks of birds are able to organize themselves to fly together through incredible distances while minimizing loss of members and collision. Recently, a number of successful heuristic guidelines have been inspired by simply mimicking natural behaviour of ants, schools of fish, and even evolutionary progress observed in bacteria. These techniques have been successfully applied in a variety of problem solving capacities for the past two decades in the field of swarm intelligence, or swarm robotics when applied to guide the behaviour of physical robots
Project description
This project applies the success of swarm robotics to search and rescue scenarios. Mirco aerial vehicles (MAVs) are used to conduct distributed search in a 3-dimensional space for range-limited WiFi signals from multiple rescue targets. Once a rescue target is found, the swarm autonomously reorganizes itself to form a relay network that allows communication between the victim and the base. These MAVs are minimally equipped with basic sensors to simulate crude proprioceptive awareness. A hybrid evolutionary PSO (particle swarm optimization) framework is used to govern the emergent swarm behaviour. In the figure below, each darkened circle represents a target and its communication radius. Planes in the figure represent a swarm of MAVs and their communication relay range. As seen in the figure, the first state initializes randomly placed search targets and defines a bounded search space. In the second state, or phase one, swarms of MAVs are deployed to match the number of targets. The swarms spatially form kin groups and divide up the search space into separate territories. However, as the search continues, migration from the original search space may occur. Since spatial relationships determine membership to kin groups, exchange of members between kin groups is possible. This encourages solution diversity within kin groups, which may improve performance. The second phase is triggered when a target is discovered. The MAV that first located the target will calculate its position from base of operations by reviewing its flight history. It will then send control signals to recruit other MAVs in its vicinity to form a relay network.
Design
methodology
-
Extensive
review
of
current
literature
on:
- Conditions facilitating emergent communication
- State-of-the-art solutions to similar application scenarios
- Design and simulate control algorithm using hybrid PSO and evolutionary strategy
- Analyze emergent behaviour and iterate process to seek improvements