18 October 2016
According to Baidu, one of the oldest testers of autonomous technology, self-driving cars still have a long road ahead of them. There are still many obstacles to be dealt with before the cars can be released to the public. In a recent report Baidu goes deeper into some of the current technical weaknesses of autonomous technology.
Baidu (aka the Google of China) has been testing its own self-driving cars for many years. The company aims to release their first public models to the market around 2018. In the process they have found many obstacles, for some of which they still have not found a solution. These findings would represent the weaknesses of autonomous technology for the entire driverless industry. Luckily Baidu accompanies each concern with a solution.
Emergency situations, construction works and accidents
At the moment, self-driving vehicles have no way to communicate with emergency vehicles. Ambulances and fire trucks for instance. Their unpredictable road movements at high speeds cannot be combined with basic road scanning. Autonomous cars must communicate with them on a different, premeditate channel. Unfortunately, not a lot of manufacturers seem to be looking into this.
Also in case of road constructions or accidents, self-driving cars must have a way of interacting with construction workers or traffic police, since the vehicles cannot interpret hand signals for example Baidu suggests that the best way to solve the issues is by developing an overarching application that would tell the autonomous car what to do/where to go in case of these special situations.
Self-driving vehicles already have a way to interact with each other and understand traffic signals. Usually this is by taking photos and interpret them (through deep learning). Here the biggest malefactor is glare. A blinding ray of light can interfere with the quality of the photos, and cause the car to misinterpret the signals, just like it effects a human driver. And we all remember what happened recently in Florida with a self-driving Tesla.
The solution Baidu comes up with is investing into adding more traffic signals on our roads (and adjust the existing ones). More signs, facing different angles could avoid the chance of glare affecting the quality of photos.
The visibility of road signs
Together with glare comes the issue of visibility of road signs. Although this weaknesses of autonomous technology can be considered a minor issue, it is very important that road marking are extremely visible in order for self-driving vehicles to scan and understand them correctly. The same goes for lane markings. The slightest smudge or damage could cause the cars to misinterpret the signs. Here again Baidu emphasizes the importance of patching up our existing our current traffic signals or maybe even upgrading them.
News tags: Baidu,