The trend towards autonomous mobility, in all sectors, is “unstoppable” leveraging on several technologies that have been evolving in the last decade and that are promising to become affordable in the current one. Automation in mobility feeds on the same forces that pushed automation in manufacturing: better and consistent performances and lower human involvement (that implies lower risk of harm to human operators, lower cost in operation and training, higher flexibility, no distractions nor tiredness…).
The reasons why we have been waiting for so long in pursuing autonomous mobility is quite simply that it is much more difficult to manage an ever changing situation (as you move around the context is bound to change) and it requires a higher degree of awareness on the overall situation plus on the evolution of the situation, as actions are taken by anybody in that context.
Advanced sensors, signal processing and artificial intelligence are the crucial ingredients (and of course effective actuators!). This also requires quite a bit of processing power, more than was available (at an affordable cost) only few years ago. The processing in principle could be located outside of the autonomous “moving” device but in that case the latency in the communication channels (both the one connecting the device to the processing point and the other from the processing point to the device) should be kept below a given thresholds, depending on the situation (the type of environment and the speed of the device…).
This is the reason why the (potential) low latency of 5G and the deployment of edge computing and edge cloud to bring processing closer to the final user is so interesting.
However, the advance in processing performance, at affordable cost, is now co-locating the required processing inside the device, leaving to the edge cloud/computing the role of global/local awareness and supervision (like to understand the overall traffic in a given area). In perspective, early next decade probably, we might have so much processing power within each single autonomous device and a pervasive device to device communications fabric that the need for an edge computing may relent (welcome to 6G).
The first approach to remove the humans from the wheel is based on inventing something that can replace the human, like the PIBOT robot (watch the clip). At KAIST researchers have created a robot that can sit at a pilot seat and fully control the plane from take off to landing. It has been designed with the goal of replacing a pilot for missions where a pilot would not be able to operate the plane (imagine a nuclear accident where a pilot would not be able to get close to the accident site because of radiation). Previous versions were designed to help the pilot acting as “second officer”.
The interest in this approach (having some artefacts impersonating the pilot at the wheel) has the advantage that it could be applied to existing vehicles (you just need to design it in such a way that it can fit). However, if you can start from scratch the best approach would be to “remove the wheel”! This is what the FTI’s report predicts for the coming years and ti will be considered in the next post.