Swarm coordination with robust control lyapunov function : formulation and experiments

• Main Author: Chiew, Soon Hooi
• Other Authors: Go Tiauw Hiong
• Format: Theses and Dissertations
• Language: English
• Published: 2014
• Subjects: DRNTU::Engineering::Aeronautical engineering
• Online Access: https://hdl.handle.net/10356/55326
• Institution: Nanyang Technological University

Flocking and swarming are inspired by nature. Swarming formation involves a large number of agents or unmanned aerial vehicles (UAVs) equipped with basic sensors or payloads. Missions are completed as a result of emergent behavior as a whole, relying on local sensing and reactive behavior. The individual position of UAVs in the swarm is not as important as in a formation flight, but rather the overall cohesion, separation and alignment requirements are the more important aspects. A decentralized scheme with behavioral approach is deemed the most feasible for swarm coordination due to its close resemblance with flocking in nature that enables an intuitive modeling of swarming behavior. The inter-agent interaction is described by artificial potential field based on topological distance. Robust control Lyapunov function (RCLF) approach is attempted to formulate the swarm control as a decentralized robust stabilization problem. It remains as a challenge to model system uncertainties in swarming control, such as those originated from measurement, telemetry, external disturbance or the nominal plant itself. These uncertainties could pose detrimental effect to the swarm. These problems are further complicated by the number of agents that presents in a swarm. The coordination strategy in this research addresses these challenges and has several advantages. The scalability of the swarm is achieved with decentralized architecture that allows flexibility in altering the swarm size. This enables the attrition and expansion of the number of agents. The swarm is also robust to individual agent’s failure or imperfection because the malfunction of an agent or a leader will not pose any danger to the entire group. This is because any agent can take over the leader’s role should the leader fails, without informing the other agents about this change. Simulation and experimental results are presented to show the feasibility of the approach. The experiments were first carried out with commercial quadrotor platform. In order to have more customization and flexibility on the test platform, a custom quadrotor was developed based on Open Hardware Open Software projects. The performance of the custom made quadrotor is presented to show its comparable capability with the commercial platform. An experiment with four custom made quadrotors was performed for a three dimensional scenario in the motion capture laboratory using the Vicon system.