Autonomous driving research has been conducted almost exclusively in environments with lane markings, predictable pedestrian behaviour, and comprehensive regulatory frameworks. African road infrastructure operates on entirely different terms. Kampala, Nairobi, and Lagos feature mixed-use roads where motorcycles, pedestrians, livestock, and heavy trucks share space governed by informal norms rather than traffic signals. Building a system that performs reliably here requires different assumptions at every layer of the stack.
This project focuses on three areas: sensor fusion architectures that perform reliably in high-dust, high-glare, and inconsistent lighting conditions; edge inference models that can run on affordable hardware without cloud connectivity; and a regulatory engagement framework for jurisdictions that have no existing autonomous vehicle legislation.
The sensor fusion work addresses a specific problem: the LiDAR and camera sensor combinations validated in North American and European testing environments degrade significantly in the dust, glare, and unpredictable lighting conditions common to East African urban roads. Our research is evaluating alternative sensor combinations and fusion architectures that maintain reliable object detection under these conditions without requiring sensor hardware priced beyond the economics of the market.
The edge inference component addresses connectivity constraints. Autonomous systems designed for markets with reliable high-bandwidth connectivity can offload compute-intensive inference to cloud infrastructure. In environments where connectivity is intermittent or expensive, the inference must run locally on hardware with meaningful cost constraints. We are benchmarking edge inference platforms against a driving scenario dataset constructed from Kampala traffic conditions.
Current status: completed environmental modelling for Kampala's central business district, benchmarked three edge inference platforms against our driving scenario dataset, and initiated engagement with the Uganda National Roads Authority on a regulatory sandbox proposal.
