Mar 29 2018
A global research project is focused on developing a pioneering new collision avoidance system to improve the safety of driverless cars. The project derives its inspiration from swarming insects.
Using a €1.8 million grant from the European Union's Horizon 2020 research and innovation program, the project will build a tiny, trustworthy collision detection sensor system that could significantly advance the safety of autonomous vehicles.
Although broad testing of these vehicles has already started both on and off road, their safety around other vehicles and unforeseen hazards is a major stumbling block in their progress.
The ULTRACEPT project - which stands for ‘Ultra-layered perception’ with brain-inspired data processing for vehicle collision avoidance and is led by the University of Lincoln, UK - will create a new microchip for driverless vehicles which aims to make them safe to enter human society.
Developers have found that the existing approaches for vehicle collision detection are essentially ineffective in terms of cost, reliability, size, and energy consumption: radar is too sensitive to metallic material, GPS-based approaches face hurdles in cities with high buildings, vehicle-to-vehicle communication cannot spot pedestrians or any unconnected objects, and normal vision sensors cannot handle rain, fog, or dim light conditions at night. The ULTRACEPT researchers hope to build a system that will overcome all of these matters.
The new ULTRACEPT sensor will draw inspiration from the rapid reactions of insects, integrating long-range hazard perception, near-range collision detection technology, and thermal-based collision detection tools. This will guarantee that it functions day and night, and can swiftly adapt to unforeseen hazards and varying conditions - for instance driving in and out of tunnels or unexpected weather changes.
This means that the robust, economical, and energy-efficient collision detection and avoidance system will deliver a capability which is presently beyond the autonomous vehicles in development.
The project brings together specialists from universities in the UK, Japan, China, Germany, Malaysia and South America. Professor Shigang Yue, Professor of Computer Science at the University of Lincoln, is leading the ULTRACEPT project.
He said: "Autonomous vehicles, although still in the early stages of development, have demonstrated huge potential for shaping our future lifestyles - from sending children to school, driving commuters to work, delivering packages to households, and distributing goods to warehouses, shops or remote areas. But to be functional on a daily basis there is one critical issue to solve; trustworthy collision detection.
"Biology provides a rich source of inspiration for artificial visual systems for collision detection and avoidance. For example, locusts, with a compact visual brain, can fly for hundreds of miles in dense swarms free of collision; praying mantis can monitor tiny moving prey with the help of specialized visual neurons, and nocturnal insects successfully forage in the forest at night without collision.
"These naturally evolved vision systems provide ideal models to develop an artificial system for collision detection and avoidance, and we hope that in the future, each vehicle, with or without a driver, will be well equipped with an innovative sensor to navigate as effectively as animals do."
The project will be undertaken by a world-class research team that will include experts in hardware and software systems and robotics, invertebrate vision modelers, mixed-signal chip designers, invertebrate visual neuroscientists, robotics platform providers, and brain-inspired pattern recognition.
It will be based on Professor Yue's expertise in developing autonomous navigation of mobile robots that drew inspiration from the locust's unique visual system, as well as the work undertaken as part of the earlier 'Spatial-Temporal Information Processing for Collision Detection in Dynamic Environments' (STEP2DYNA) research project, also guided by the University of Lincoln.
The University of Lincoln is working with Hamburg University and Newcastle University for STEP2DYNA, plus partners from the University of Buenos Aires in South America, Kyushu University in Japan, and Chinese institutions, Huazhong University of Science and Technology, Xi'an Jiaotong University and Tsinghua University.
Joining the consortium for ULTRACEPT is the University of Münster, Universiti Putra Malaysia, National University Corporation Tokyo University of Agriculture and Technology, Institute of Automation Chinese Academy of Sciences, Lingnan Normal University, Northwestern Polytechnic University and Guizhou University, plus SMEs Visomorphic Technology in the UK, and German-based Dino Robotics.
Most autonomous vehicles are still legally restricted to their testing arena precisely because developers can't be sure of their safety and accidents involving Tesla and Uber cars, with autonomous driving functionality on, have only highlighted this issue further. Collision detection and avoidance is so important for vehicles now and in the future, yet there is no acceptable product currently available on the market to specifically meet this need - that is exactly what we hope to develop.
Professor Shigang Yue
The project will also involve designing an all-inclusive video database of risky driving scenes so that the new systems being developed can be thoroughly tested before being launched onto the road.
Besides being used within autonomous vehicles, ULTRACEPT's integrated computer vision system will be relevant to a number of other industries, including video game developments, robotics, and healthcare.