SwarmLab-Projects

We welcome students to work on swarm projects using our simulation framework or the real robots. This page shows the currently active thesis that we are supervising. If you want to take on one of the proposals or have an interesting idea yourself please contact us at steup-at-ovgu.de.

 Wireless Ranging between Copters 

The sensory equipment of our sensor has shown some limitations. Especially the sonar ranging sensors cannot reliable detect other copters. The copters' construction does not reflect incoming sonar waves good enough to provide a reliable estimation of distance between copters. Therefore, other sensory systems need to be evaluated. Wireless ranging may provide a more robust ranging system between copters. This Bachelor Thesis aims to integrate a Wireless Ranging System in our copters to provide an additional sensory input and a communication link that is more flexible then the currently used bluetooth system.

 

Realistic Simulation Model of Finken II 

The current simulation uses a very rough estimation of the physical properties of our copters. To be able to transfer the results of our simulations to the real swarm we need to enhance the realism of the simulation. This includes a non-linear model of the copters behaviour to different internal and external disturbances and control inputs. The model developed in this Bachelor Thesis will then be included in our Mixed Reality Simulation Environment to provide more realistic results.

 

Simulation: Effects of Environment on the Swarm Behaviour 

Our current simulation can physically simulate the copters and evaluate simple swarm behaviour like cluster formation. In the next step we want to check the robustness of the behaviour of the copters to various environmental influences like wind, changes in air pressure or failures of sensors. The goal of this project is to model such envioronmental influences in the simulation together with the influence on the copters. Finally, the project shall investigate some swarm behaviour algorithms for cluster formation against different environmental disturbances.

Simulation: External control of Swarms of Copters 

Currently our virtual copters swarms randomly walk through the arena without a goal. This project will enable the simulation to set goals to the copters using virtual landscape controlling the behaviour of the copters. This enables new formations the copters will follow as well as constant movement if the landscape is constantly changing. Existing work on integrating dynamic landscape in the simulation and detecting these with the copters can be used. The goal of this project is to evaluate the necessary sensory equipment and the control parameters necessary to move the copters stable and without collisions.

Advanced Control API of the Copters 

Currently, our copters are directly controlled using the pitch, roll, yaw angles and the thrust. However, this control is very low-level and not very good suited to our high-level swarm algorithms. This projects aims to provide a more indirect control of the copters using other control inputs like speed or positions, which eases the implementation of the swarm algorithms. To this end, the sensory input of the copters need to be extended to provide at least estimates of speed and positions. Afterwards control algorithms eed to be developed, which uses these inputs and the set points of the application to actuate the copter using the low level controls pitch, roll, yaw and thrust.

Robust Multi-Copter Flight 

Most of the work until this point aimed to provide a single copter with stable behaviour,  swarm algorithms could rely on. This project focuses on more than one copter with the goal to provide a stable behaviour. The goal is to let a minimum of two copters fly in the arena without collisions to each other and the walls. The existing work on wall-avoidance and height control are used to implement baseline behaviour. The difficult part is the robustness, since sensory input are unreliable and the copters will influence each other with the wind they generate.

 

Multi copter flight with one fully autonomous and one manually controlled copter
Robust Take-Off and Landing (Software Team Project):

Currently our copters cannot use their autonomous behaviour during take off and landing. This is partially because the used sensors are outside of their operational range and partially because the autonomous behaviour is based on a free range around the copter. A third aspect is the change of behaviour of the copter due to the ground effect. The aim of this project is to analyze the existing autonomous behaviour for height control and wall avoidance during take off and landing and modify it to be able to have a very robust take off and landing.

 
Old start and landing behaviour
 
Optimized start and landing behaviour

 

Avoiding Walls with Quadrocopters

 Our FINken copters shall fly robust and safe in our arena in G29-035. This is the baseline functionality needed to implement higher order swarm behaviour. To this end, it is necessary to evaluate the performance of the used sensors and adapt existing algorithms like attraction and repulsion to avoid the walls. Focus in this topic is the evaluation of the sensors and the robust behaviour of the copter.

Short Video of Wall Avoidance

Longer Video of Wall Avoidance

Cooperative Control of Height

This topic aims to provide a first showcase of swarm behaviour. Currently our copters are able to control their hight autonomously, but only with a statically coded hight value. In this task 3 are fixed in test setups, so that they are only able to change their hight. The copter shall react to external interference and change their height accordingly in a swarm.

Test System for Autonomous Height Control 

With the new test system, we analyze the stability and other features related to the communication systems of our robots.

Test system

Stability analysis in dynamic environments

 Robustness analysis

 

Control Theory based Height Control and Wall Avoidance (Bachelor Thesis)

For a reliable experiment using the copters, a robust and reliable base control layer is necessary. This layer shall prevent the copter and the humans working on it from damage. Therefore, we envision the copter to actively avoid walls even if the swarm algorithm tells them otherwise. Additionally, it is necessary for the copters to maintain a specified height. The design and parameters estimation of control systems providing such reliable height control and wall avoidance is the goal and work of this bachelor thesis.

The thesis is finished and provided very good results, as visible in the below videos. It is avilable in German here.

 
Height Control Experiment  
Wall Avoidance Experiment

 

 

 

Last Modification: 14.06.2019 - Contact Person: Webmaster