MIT strives to develop an AI model that can drive almost like a human

Creating unmanned vehicles capable of reasoning as well as handling human-like situations is one of the most important and long-term goals of companies that are pioneering in the vehicle industry. like Waymo, GM, Cruise, Uber and some other big names. Mobileye, a well-known enterprise in vision technology development for Advanced Driver Assistance System (ADAS) and self-propelled driver, belongs to Intel Corporation, previously proposed a mathematical model with name Responsibility-Sensitive Safety (RSS). This invention is described as an innovative way for the car to make more accurate decisions during operation, similar to the way we relocate normally, and has received a lot of reviews. Positive price from the community. In addition, another big man who is also interested in self-propelled transportation is Nvidia who is also actively developing Safety Force Field - a 'policy' that helps the car make accurate decisions based on on the ability to plan to monitor and evaluate all unsafe actions during operation through the analysis of sensor data obtained in real time.

MIT strives to develop an AI model that can drive almost like a human Picture 1

1. Bosch introduces sensor boxes designed for unmanned 'flying taxis'

The above is a good example of the praiseworthy efforts of businesses in focusing on developing the ability to drive safely - the core factor that determines the success of a self-driving car. However, scientists at the Massachusetts Institute of Technology (MIT) have a much bigger ambition, to build an AI system that is capable of driving exactly like humans. More specifically, a group of MIT artificial intelligence experts are currently working on ways to leverage GPS data and other visual data like GPS to allow AI-based self-propelled cars to be available. can learn how to drive people, and at the same time apply 'fluently' the trained knowledge to the test routes, which are complexly planned in the environment never seen before. Accordingly, this study was built on the basis of end-to-end navigation systems designed by Daniel Rus, director of the US Artificial Intelligence and Computer Science Laboratory (CSAIL). , and is expected to be presented at the International Conference on Robotics and Automation in Long Beach, California, USA next month.

MIT strives to develop an AI model that can drive almost like a human Picture 2

1. The safety features equipped with Tesla cars make it difficult for you to 'want' in the wrong lane

The previous machine learning model by Daniel Rus and colleagues is considered 'cumbersome' and less likely to be deployed in depth. However, in this new model, MIT scientists have overcome almost the above problem. 'With our system, you do not need to train the car on every road, but you can download a new map for the vehicle to help the vehicle be able to navigate through the roads it never before, the engineer Alexander Amini, the head of the research team, said.

As Amini and his colleagues explain, their AI system is able to observe and learn how to control the vehicle from the driver, and then correlate the rotation of the steering wheel with each curve. It is observed through the camera system and the map that was previously included in the system. Finally, the vehicle can give a precise 'command' for each different road, such as a straight line, a four-way intersection or a fork, revolutions .

MIT strives to develop an AI model that can drive almost like a human Picture 3

1. 55 self-propelled cars come from 5 major automakers that are being tested at the same time

In many experiments, the researchers provided the machine learning model with a map with a randomly chosen route. During operation, the system extracted intuitive features from the camera, allowing it to predict each element in the structure of the road such as signs or speed bumps, hard dividers . More moreover, the AI ​​model also has the ability to correlate visual data with map data to indicate the irrationality, which helps to more accurately determine the position of the vehicle on the road and ensure the vehicle is running on the route. The safest road. For example, when the AI ​​model is controlling the vehicle on a straight road and has absolutely no turn, however, the map system has errors and indicates that the vehicle needs to turn right, the AI ​​has realized This is unreasonable and knows how to control a car in the right direction without following the order from the map.

MIT strives to develop an AI model that can drive almost like a human Picture 4

1. Is humanity on the right track in developing self-propelled vehicle technology?

'In the real world, sensors cannot always achieve perfect precision. However, we want to ensure that our model can 'adapt' to the various errors that occur in operational practice, by building a system that can recognize anomalies and remain have the ability to navigate as well as accurately locate all roads, 'added engineer Alexander Amini.