CEDoc-UM6P-AI MOV : Event-aided Visual Navigation for Agriculture Full-time

Mohammed VI Polytechnic University is an institution dedicated to research and innovation in Africa and aims to position itself among world-renowned universities in its fields

The University is engaged in economic and human development and puts research and innovation at the forefront of African development. A mechanism that enables it to consolidate Morocco’s frontline position in these fields, in a unique partnership-based approach and boosting skills training relevant for the future of Africa.

Located in the municipality of Benguerir, in the very heart of the Green City, Mohammed VI Polytechnic University aspires to leave its mark nationally, continentally, and globally.

1 Research motivation Visual navigation, or computer-vision-based navigation, is the task of guiding a robot towards a goal using a camera as the main exteroceptive sensor. The use of a camera in navigation, especially aerial navigation, is very appealing due to its low cost, light weight and compactness. Visual navigation is an old problem in computer vision and robotics research communities. Successful approaches have been applied to both aerial and ground robot navigation[2, 1]. One of the major challenges in visual navigation is robot localization errors which may lead to robot drift over time. Such errors could be due to illumination changes between two camera frames, motion blur, etc. In this thesis we want to address the visual navigation problem by leveraging data provided by a second sensor: event camera.

2 Problem statement Visual navigation poses significant challenges on its own. In recent years, event cameras have gained increasing interest in the field of robotics [5, 4]. Unlike standard cameras that capture images at a fixed rate, event-based cameras asynchronously measure changes in per-pixel brightness, referred to as events1 . Event cameras, such as the Dynamic Vision Sensor (DVS) [3], offer several advantages over traditional cameras, including: (1) a high dynamic range of 120 dB compared to 60 dB for standard cameras, (2) high temporal resolution with sub-millisecond precision, (3) low power consumption at 23 mW, and (4) minimal motion blur. These properties make event cameras particularly suitable for challenging conditions where standard cameras may struggle, such as outdoor navigation with varying lighting conditions, highly illuminated environments, highspeed navigation, and long-term operation. Such scenarios are common in agriculture and need to be effectively addressed.

This thesis aims to explore the impact of event cameras on visual navigation in agricultural contexts. Specifically, we seek to determine whether events provide significant improvements in localization accuracy and to what extent event-aided navigation can reduce robot drift over time.

3 Research scope To answer the aforementioned questions, the candidate will first implement a state-of-the-art visual navigation algorithm. Tests will be conducted with Unmanned Aerial Vehicles (UAVs) and Autonomous Ground Vehicles (AGVs) in both simulated and real environments. Additionally, experiments in challenging scenarios will be carried out with and without the event camera to evaluate its impact on navigation.

4 Admission criteria The PhD position is proposed by the International Center of Artificial Intelligence of Morocco, of the Mohammed VI Polytechnic University. Applicants with excellent curriculum must be holders of a Master, an engineering or an equivalent recognized degree in Computer Science or Applied Mathematics. Experience with 3D computer vision and robotics is desirable but not required. In addition, they should have skills in Programming (Python and C++) and good communication skills in English. Particular attention will be given to the suitability of this research project with the applicant’s background.

References

[1] Suman Raj Bista, Belinda Ward, and Peter Corke. Image-based indoor topological navigation with collision avoidance for resource-constrained mobile robots. Journal of Intelligent & Robotic Systems, 102(3):1-24, 2021. [2] Andy Couturier and Moulay A Akhloufi. A review on absolute visual localization for uav. Robotics and Autonomous Systems, 135:103666, 2021.

[3] P Lichesteiner, C Posch, and T Delbruck. A 128× 128 120 db 15µsec latency asynchronous temporal contrast vision sensro. IEEE Jornal of Solid-State Circuits, 43(2):566-576, 2008.

[4] Yansong Peng, Hebei Li, Yueyi Zhang, Xiaoyan Sun, and Feng Wu. Scene adaptive sparse transformer for event-based object detection. arXiv preprint arXiv:2404.01882, 2024. [5] Nikola Zubic, Mathias Gehrig, and Davide Scaramuzza. State space models for event cameras. ‘ arXiv preprint arXiv:2402.15584, 2024.

UM6P.