This article aims to present a design for training AI based self-driving cars. Look this should had been filed as a Design Patent technically, but I don’t have enough funding. So in case you want to enhance it, please do give this article in your reference section. I have explained it all in form of question and answers.
Ques. What was the reason to propose Learning AI enabled Simulation Fields for self driving cars?
Ans. Its have been a while AI has developed well to advanced stages, but why Google Self Driving AI based cars are yet not out in news much ?
Ques: Why do we need AI enabled self-driving cars ?
Ans: Well, my concern to develop self driving cars is to help in areas where humans cannot drive in, example in situations such as forest fires. There can be technically many solutions for forest fires, via air or via land. Both are equally essential, for several reasons. In case of stopping forest fires via land, many times fire fighters fail to get in, given the scale of forests sizes and intensity of fires. Here, fire proof, self driving AI based car/trucks can help in to tackle the forest fires. This is one example, I can provide more illustrations if required.
Ques: How do a typical AI enabled self driving car operate?
Ans: This theoretically should use, sensors, computer vision, acoustics and learning rules with some driving lessons fed in the architecture. Knowing the safe distance form any other object of road, integrating with Google Maps to find the target locations is just some of the trivial things a self driving car is expected to do. Apart from that real time image processing in higher dimensions to know of all obstacles, to read in traffic signals, to trap traffic light signals, to automatically control speed of the AI based car using traffic analysis both live on road and as per GPS and Google Maps predictions. These are some of the basic AI based functions that should be there, theoretically, in any AI based self driving car.
Ques: What is this design for AI enabled simulation fields for self driving cars all about ?
Ans: In this article I present the AI enabled self driving car simulation pads in an interactive question answer form. The simulation pads as proposed here in this article are the learning fields for the self driving cars, where in full field is allocated to self driving cars. The field is itself AI enabled with lot of AI entities doing robotic works. Lot of simulations needs to be performed by AI enabled self driving cars, before they can come out of labs to real life. These simulations include not just sensing an obstacle and taking sides, learning to turn on right times. No, these simulations shall be like auto-pilot simulation training for AI enabled self-driving cars.
Ques: Where is AI in these self-driving cars simulation fields ?
Ans: The AI simulation labs shall be fully AI based, all robotic entities shall be present, with sensors at all key points to measure the performance of the self driving car we making learn. All these are learning points, real time feedback shall be given to the self driving cars on all entries, be it speed, be it missing a robotic traffic light operating on right intervals, or be it honking at wrong places. For each mistake the car does a feedback is provided to make it improve the learning, one the deep learning network is learned for this 5 sq. km. area. The self driving car can be send to the evaluations and testing units.
Ques: How big the simulation fields need to be?
Ans: Well It should be at least 5 kilometer square in area. To include man-made AI identities of all entities that are present in a colony and roads around it. This shall not be just sensors and roads, but lanes, trees, parking lots, markets, artificial robots performing simulations to act as humans in between the area. Apart from that it should have robotic animals moving in to various places. This shall ensure even animals are protected by the AI based self-driving cars. Further, other robotic cars shall be working in this simulation fields for self driving cars. However, these need not be AI enabled, just robotic, to work as per robotic work on predefined paths.
Ques. How does evaluation and testing of self driving car be performed ?
Ans. Well when the error is minimum, and it is considered that the car is ready for evaluations. The same AI simulation lab shall load some random test data entries, which may be stored on the grid. This can be compared to as driving license test for a person seeking to drive a car. This driving license for self driving car, can be provided if it passes several predefined tests loaded on the same simulation field. Once it fails, it is send back for training and learning as explained below in the AI based simulation labs. If the self driving car passes these AI simulation tests, it is send to next level — the next AI simulation field in another location. This is done to make it learn more, to make it more robust. Imagine the deep learning network it learns after all this. Can you imagine, as all cannot be rule based.
Ques. Where else can these AI simulation fields to test self driving cars be used?
Ans. These AI simulation fields in locations can be rented to allocated automated driving licenses to people, seeking the same. How, as all sensors are fixed at right intervals, all real time AI based entities are working and real time driving shall be tested with this, this shall not be just checking how to reverse the car, and fit it in parking lot but all the possible things. This is a big application of use of AI simulation fields in free time. Doing this with help of human evaluator shall also help improve AI simulation fields more.
Ques. How can AI self driving car help in forest fires ?
Ans. A simulation field of forest need to be created, the real time human based monitoring need to be done to tell the car, where to emit the water streams to. These wont be normal cars, but perhaps like thinner trucks which can sandwich in forests, saving any animal on the way — -and even cutting the trees in order to move forward with its water tankers, either to save life’s or to start the water supply it has brought with it in connecting pipes. This shall be much easier to learn, as it can be jointly run with help of humans sitting remotely, getting all visual images from the place the car is driving to.
For now, this is it. In case you are tweaking it, do give a reference to this article. Thanks for the support.