Currently, the world is experiencing a paradigm shift in the area of autonomous driving. The automotive industry is moving from partial driving assistance (Level 2) to conditional driving automation (Level 3). There are multiple obstacles to deploying level-3 autonomous driving, such as (lack of) regulations, business models, technology, and – quite important – costs. Nowadays, level-3 cars require the use of a high number of sensors such as cameras, radars, and LiDARs, which need to cooperate to have a very detailed knowledge of the surrounding environment. The use of these sensors makes the cost of the car increase drastically. Among all, the impact of LiDARs (Light Detection and Ranging) is the one affecting the most the price of the car sold to the final users. Indeed, a 3D LiDAR, a sensor that provides a high-resolution 3D view of the car’s surroundings as it goes, costs several thousand dollars.
The SMARTCARS project aims at abating the costs while still guaranteeing reliability by replacing 3D LiDARs with 2D LiDARs: small, power-efficient, and cost-effective sensors, with an average price of 200 dollars. This can be achieved by a cooperative approach, merging the information retrieved by multiple LiDARs in order to get a comprehensive understanding of the surroundings. The idea is not to simply mount multiple 2D LiDARs on the same vehicle and merge their data, but instead to exploit the features provided by the 5G technology to share real-time data with LiDARs of other cars.
The system we designed consists of a high-level architecture that accounts for the steps needed to make the data valuable and interpretable by each vehicle. Using machine learning and signal analytics tools (DBSCAN, SLAM, Hough Transform), the SMARTCARS team aimed to develop a solution that could break down the economic constraint represented by the high cost of 3D LiDARs, as well as pay more attention to an undervalued aspect of technological research in this field, namely that of cooperation between several vehicles through the mutual exchange of sensor data. Since we obtained some positive outcomes from the developed system, we can assert that our solution is technically feasible and works in specific and controlled situations. We then outlined how the project could be improved and developed so that external institutions could benefit from all the work done.
Principal Academic Tutors
Umberto Spagnolini, DEIB, PoliMI
Academic Tutor
Marouan Mizmizi, DEIB, PoliMI
Dario Tagliaferri, DEIB, PoliMI
External Institutions
SMARTCARS, Huawei
Team members
Pietro Macori, Computer Engineering PoliTO
Monica Moser, Mathematical Engineering PoliTO
Paolo Motta, Mathematical Engineering PoliMI
Giacomo Preti, Automation & Control Engineering PoliMI
Francisco Javier Sánchez Olivera, Aeronautical Engineering PoliMI