Expectations on the performance of self-driving cars are high. Very high. To pass the test of social acceptability, they must do more or less better than humans. To achieve this, the industry is therefore striving to develop a panoply of sensors, such as cameras, GPS systems, radars and lidars. Then, you have to put all the data collected together in a software to be able to process them in real time in order to have a precise idea of what is happening around the car. The company LeddarTech develops this kind of software, in addition to lidar sensors that use a light signal to detect objects.
“We work with the client’s specifications. Generally quite accurate, he may mention for example that we need to develop a sensor capable of detecting a pedestrian at 200 meters, with a volume of less than 200 cubic centimeters, which costs less than $200 to produce, which has a lifespan on a car that has been used for 25,000 hours and which operates from -40 to 105 degrees Celsius,” explains Pierre Olivier.
Software development is also a big challenge. LeddarTech actually creates a three-dimensional environmental model in real time that is updated between 10 and 30 times per second, depending on the customer’s needs, by processing data from the different types of sensors present on the cars.
“We have two cars in Quebec and two in Israel that are equipped with different sensor systems, and each of them collects in one hour more than two terabytes of data, which normally would take days to transfer, illustrates Charles Boulanger , CEO of LeddarTech. But the software must make instant decisions, without making mistakes. Robustness in the transmission of information is essential. »