Smart bikes: trajectory prediction with onboard sensors

Smart bikes: trajectory prediction with onboard sensors

Problem statement

Trajectory prediction is essential to make future bikes smart. It enables the development of variety of assistance systems for cyclists, such as lane keeping, safe navigation, potholes/obstacles detection.  Many sophisticated methods have been proposed for autonomous vehicles. However, those methods might not be feasible for bicycles, since bikes are cheaper and have limited space and resources (computing power, memory, etc.). It brings additional challenge to percept the environment and interpret them in a lightweight manner.

Task

You have a chance to work with a group of researchers and industrial partners to pioneer the work in this field. We plan to use IMU sensors and low-resolution cameras on the bikes, collect data in real time, estimate the position and predict bike’s future trajectories. The feasibility of available methods needs to be studied and a new solution dedicated to bikes will be developed. Depending on your interest, you can either review existing methods to make a survey or improve existing methods to make a new state of the art. All facilities, such as (E-)bikes, sensors and processors will be provided.

Work

40% Theory, 40% Programming, 20% Writing 

Contact:

Yanqiu Huang, yanqiu.huang@utwente.nl