The robotics lab is part of an ambitious project that aims to create a new teaching and research space in which bachelor, master, and doctoral students can find different types of robotic platforms in order to test in embedded systems ideas and concepts developed in different teaching modules. The Lab features:
- 10 e-puck robots each of which equipped with hardware extensions such as the range&bearing board and the pi-puck module;
- 50 kilobot robots;
- 20 in-house made robots named choir-bots;
- two arenas to run robot experiments.
A 2-year Post Doc position open in this Lab. More details here.
The majority of the research work carried out in the Lab concerns problems related to the design of control mechanisms underpinning individual and/or group level adaptive processes in autonomous robots. We are interested in developing the mechanisms allowing the robots to autonomously adapt to complex and varying social and physical environments. We pursue this objective by using different design methods based on a collections of AI and machine learning techniques.
Weed detection with UAVs
Agricultural areas are very large and weeds are tedious to be discovered. In the context of precision agriculture, a swarm of micro-UAV (MAVs) is deployed to autonomously and quickly fly over the crops to analyze the plants in a non-destructive way. A deep learning network (Faster-RCNN) embedded on the MAVs analyzes the images taken by the MAV to detect weeds in a sugar beet field and then compute a statistical infestation map of the crops. The swarm organization shares the information collected by each individual to the whole swarm to establish a generalized knowledge base. The objective is to precisely locating each weed to enhance weed control strategies. A local spraying reduce the use of pesticides and contributes to a more sustainable and ecological agriculture.
Collective decision making in a swarm of robots
Swarm robotics is a field of research that studies how systems composed of multiple autonomous agents can accomplish collective tasks. The swarm is characterized by distributed control, simplicity of individual robots and locality of sensing and communication. Due to the distributed nature of swarm robotics and the absence of globally shared information, in some task scenarios, the coordination among robots swarm requires a process called collective decision-making (CDM). Our work aims to design a mechanism for CDM in a swarm of robots by exploiting the evolutionary robotics approach and by using the collective-perception scenario as a test bed to validate this mechanism.
In the collective perception scenario, the robots explored a close arena with the floor made of black and white tiles. The swarm has to reach a consensus on which type (black or white) of tiles covers most of the arena floor. We intend to study how the spatial distribution and number of features scattered in the environment influence the CDM process and individual behaviour. Moreover, we intend to investigate the ability of the synthesized controller to adapt to new changes in the environment. Finally, we plan to port onto physical robots (e-puck) the CDM studied in simulation. Ahmed Almansoori thanks UNamur for the financial support through the CERUNA studentship.
Auditing electrical infrastructures with UAVs
The aim of the project is to set up and deploy one or more fully autonomous UAVs to automate the inspection of high-voltage line installations such as electrical pylons and concrete poles. Currently, audits are carried out either by fitters with line logging, or by helicopter or manually piloted UAVs. The detection of defects, once the data has been acquired, still has to be done manually by an experienced person capable of analyzing and identifying the slightest defects. An enormous amount of data has to be visualized, at considerable cost.
The aim of this research project is to automate and make the inspection of towers and power lines autonomous. The pylon requires meticulous inspection of its various components (members, angles, insulators, etc.). In addition, there are several different tower silhouettes. It is therefore a real challenge to design an effective AI system capable of the pylons audits and the analysis of defects. More details here. G. Maitre thanks SPW for the financial support through the Doctorat-en-Entreprise fellowship.
Self-organised aggregation in swarms of robots
Aggregation in swarm robotics is a well-known behavior that enables the swarm to regroup before performing other tasks. We study here an aggregation problem occuring on two sites. By injecting a certain degree of heterogeneity into the system via informed individuals, we can manage the final expected distribution of the swarm on the sites. Robots are equipped with simple probabilistic finite state controllers and simulations ran using ARGoS.
Modifying the number of informed robots (robots having a preference for one of the two sites), we effectively drive the system close to the expected state and avoid the classic symmetry-breaking case where all robots end up aggregating on only one site. This research is funded by the ARIAC project, SPW Walloon region. ARIAC is an initiative of the AI institute TRAIL.
Synthetic Choir – An art/science collaborative project
This is a science-art collaborative project, created by Prof E. Tuci (Faculty of Computer Science, UNamur), Prof. T. Carletti (Depart. of Mathematics, UNamur), TRAKK (Namur), and the Belgian artists from the collective VOID. The project is about designing and building 20 autonomous robots that randomly move in an enclosed arena. The robots are made of a 3D printed chassis that hosts the batteries, the electronic and mechanical parts, and a 3D printed anthropomorphic head.
The robots can avoid obstacles using ultrasonic sensors, and they can communicate with each other using a mechanism based on infrared signals. The robots that by chance get closer to each other, they stop and start emitting a sound, prerecorded by the artists.
The continuous random interactions between the 20 robots will generate a self-organized sound and visual experience that we hope the audience will enjoy. The robots have been designed and built by the UNamur research fellow Dr Muhanad Alkilabi. This art installation will be shown in different national and international venues. When not used in the installations, the robots will be used by the Faculty of Computer Science and the naXys Institute for teaching, research and outreach activities. On 30th October 2021, the project has been presented at the RTBF radio program Les Eclaireurs .
Best-of-n problems in swarm robotics
Judhi’s research focus is in studying Collective Decision Making process, particularly in Best-of-n problem situations using bio-inspired swarm robots. Both software simulation and real hardware robots were used in this research. The research involves multiple factors influencing the Best-of-n problem process, including the presence of zealots, various voting mechanisms (voter model and majority rule), and dynamic site qualities.
- Elio Tuci (Lab Director) elio.tuci at unamur.be
- Muhanad Alkilabi (PostDoc) muhanadeng2005 at gmail.com
- Guillaume Maitre (PhD) guillaume.maitre at unamur.be
- Ahmed Almansoori (PhD) ahmed.almansoori at unamur.be
- Antoine Sion (PhD) antoine.sion at unamur.be
- Antoine Hubermont (PhD) antoine.hubermont at unamur.be
- Ziya Firat (PhD) ziya.firat at unamur.be
- Judhi Prasetyo (PhD) judhiprasetyo at student.unamur.be
- Giuliana Pagliauca (Erasmus) giuliana.pagliuca at student.unamur.be