Hypatia
trying to be a very good ENTP
- Joined
- Dec 1, 2011
- Messages
- 615
The economic race to develop the world’s first fully self-driven vehicles is currently underway. Before being approved for commercial use, it is likely that vehicles which achieve SAE levels between 3 and 5 will be considered the most viable.
Most vehicles today are manually controlled (level 0); outside of momentary warnings and minimal intervention, the driver must perform all manual tasks.
At level 1, the vehicle is capable of assisting with either steering or braking, e.g. cruise control.
At level 2, the driver is still required to keep their hands on the steering wheel at all times in case of emergency manual intervention but the vehicle itself can already steer, brake, and accelerate, e.g. lane assist + adaptive cruise control.
At level 3 the driver is able to experience sustained ease from continuous driving tasks. They can, for example, watch a movie or make a phone call, with assurance that the vehicle can automatically operate even under emergency circumstances. Some exceptions might still apply; however, in these cases, any manual assistance required will be announced in advance.
At level 4, the manual driver will be able to experience complete ease from performing all manual tasks. Driving, however, is only limited to designated areas and/or conditions. Once outside of geographical range, the driver must once again ensure that the car can safely conclude its own trip.
At level 5, the vehicle is capable of full self-guidance under virtually all road conditions.
In order to respond accurately to environmental surroundings, today’s self-driving vehicles are typically equipped with either laser sensors (LiDAR) or camera technology.
LiDAR (Light Detection and Ranging) is a remote sensing system which emits rapid pulses (or a continuous stream) of light, which then reflect off of bounded objects before they return. Transpired duration is then used to calculate distance of travel, which, when converted into elevation data, create an accurate 3D visual map.
The main drawback to LiDAR systems is that while they generate accurate environmental scanning from a distance, they lack the analogous computational power of human vision as applied to task-specific applications (computer vision).
The alternative to employing LiDAR lasers is a combination of camera and radar systems or simply a collection of video cameras which rely on neural network based computational systems that recognize surrounding objects around the vehicle.
The decisive advantage to algorithm-assisted computer learning is that decisions are made in real time, thereby making camera systems the only method which is potentially capable of achieving conditions which can approximate full self-driving.
A transportation system that relies on LiDAR must create the full gamut of visual mapping beforehand and then cross-reference that information with further test drives to ensure safety only within certain spatial parameters; LiDAR has effective applications for robo-taxis but will not be able to remove the necessity for manual labor required for a fully self-automated vehicle.
Most vehicles today are manually controlled (level 0); outside of momentary warnings and minimal intervention, the driver must perform all manual tasks.
At level 1, the vehicle is capable of assisting with either steering or braking, e.g. cruise control.
At level 2, the driver is still required to keep their hands on the steering wheel at all times in case of emergency manual intervention but the vehicle itself can already steer, brake, and accelerate, e.g. lane assist + adaptive cruise control.
At level 3 the driver is able to experience sustained ease from continuous driving tasks. They can, for example, watch a movie or make a phone call, with assurance that the vehicle can automatically operate even under emergency circumstances. Some exceptions might still apply; however, in these cases, any manual assistance required will be announced in advance.
At level 4, the manual driver will be able to experience complete ease from performing all manual tasks. Driving, however, is only limited to designated areas and/or conditions. Once outside of geographical range, the driver must once again ensure that the car can safely conclude its own trip.
At level 5, the vehicle is capable of full self-guidance under virtually all road conditions.
In order to respond accurately to environmental surroundings, today’s self-driving vehicles are typically equipped with either laser sensors (LiDAR) or camera technology.
LiDAR (Light Detection and Ranging) is a remote sensing system which emits rapid pulses (or a continuous stream) of light, which then reflect off of bounded objects before they return. Transpired duration is then used to calculate distance of travel, which, when converted into elevation data, create an accurate 3D visual map.
The main drawback to LiDAR systems is that while they generate accurate environmental scanning from a distance, they lack the analogous computational power of human vision as applied to task-specific applications (computer vision).
The alternative to employing LiDAR lasers is a combination of camera and radar systems or simply a collection of video cameras which rely on neural network based computational systems that recognize surrounding objects around the vehicle.
The decisive advantage to algorithm-assisted computer learning is that decisions are made in real time, thereby making camera systems the only method which is potentially capable of achieving conditions which can approximate full self-driving.
A transportation system that relies on LiDAR must create the full gamut of visual mapping beforehand and then cross-reference that information with further test drives to ensure safety only within certain spatial parameters; LiDAR has effective applications for robo-taxis but will not be able to remove the necessity for manual labor required for a fully self-automated vehicle.
