Autonomous driving tech is growing rapidly. The primary purpose of a self-driving vehicle is to drive without a driver. However, AD-technology is not yet perfect. While ideally a self-driving car should drive without any intervention, in current reality there will be so-called “edge cases”: situations where the AD perception stack doesn’t know what it should do. In those cases, fleet operators want to safely return their cars out of that situation. Connected cameras provide the possibility of remote car control by a human: so-called “Human In The Loop (HITL)”
Situations when a human operator is necessary can still occur. As an example, a situation where the car is driving in a route under reconstruction. At that moment all the rules the car used to follow when driving might not apply anymore. To adapt and react to these new circumstances, can be difficult even for a skilled driver, several problems might occur (an inability to detect a new lane, recognize additional signs etc.). In such cases, a disengagement with remote control would be the best solution before the car could drive autonomously again. Even though such situations may not be very frequent, it is key to solve these for safety reasons. In the USA, states like Arizona, California, Michigan, Ohio, Texas and Florida have all mandated teleoperation. Outside the U.S., laws mandating teleoperation exist in Ontario, Japan, Finland, the Netherlands and England. The remote-control solution is solving an emerging need and has grown into a business case on itself.
What constitutes remote driving?
When a car is constantly streaming its environment to the cloud, it can be remotely navigated and driven in order to make sure that the passengers and pedestrians are as safe as possible.
A remote driver sits in vehicle cockpit simulator equipped with a steering wheel, brake and acceleration pedals. Multiple screens in sync with the automated vehicle, displays the status, camera views on the road and around the car. Whenever the remote driver brakes, speeds up, or turns the steering wheel, the remote vehicle connected to the system, responds in turn. This is called “tele-operation” or “remote control”.
The car cameras are the remote driver’s eyes. One can easily understand that long (network) delays between car and remote driver, can become dangerous.
Live streaming of a (high-quality) video has always proved to be quite challenging. For a remote car with nearly 8 active HD cameras, it will demand a high bandwidth-availability from the mobile network. A delay in delivering the information is a showstopper as it directly endangers the safety of the passengers and other traffic participants. For a reliable and responsive remote control of the vehicle an intelligent and quick reduction of the video stream is required to achieve the required low latency.
Teraki addresses this bottleneck through its state of the art region of interest technique to lower the bandwidth requirements and still offering the human operator high quality images of the relevant objects. This guaranteers real-time tele-operation (and in the meanwhile also reduces the cost of data transmission). Teraki’s algorithm identifies the important areas in the live stream (such as e.g. cars, moving objects, pedestrians etc.) dynamically compress these regions retaining high visual quality. While compressing the lesser-important regions (such as sky, vegetation etc.). This is done in real time and embedded in the vehicle. By this safe remote operation is ensured - even in low available bandwidth scenario’s.
Not only for cars
While cars has the largest coverage for the remote driving operation, the same technology is equally essential for remote operation of an autonomously navigating robots such as delivery robots and of drones that fly beyond line of sight. Remote operation is a critical asset for all autonomously driven devices that still need a human supervisor until 100% autonomy has been fully achieved.
In the near future it will be a common occurrence to see driverless cars, trucks, buses, delivery robots on city streets as well as drones in the sky. A key step towards introducing autonomously driven vehicles to the public is through the development of remote monitoring and control. Low-latency remote control increases safety and hence better acceptance among traffic participants.