Transport Transformation - Is it one easy way?! (3/3) Automated driving
Driverless cars used to be a sort of science fiction, which is not the case anymore. We started to see experimental cars and trucks in a closed environments or predefined routes taking transport assignments of people and goods successfully.
In late 13th Century, Leonardo Da Vinci invented the first ever automated vehicle which is self propelled cart. It used springs and differentials like gears to travel automatically in know route. Da Vinci’s invention was ahead of his time, till mid 1920s “Francis Houdina” showed case of radio waves controlled vehicle which is the first modern view of automated vehicle. In late 1930s GM showed another case of a radio controlled electric vehicle can go by itself in a wired route. Nowadays the nearest auto maker to fully autonomous vehicle is Tesla. In commercial vehicles we saw the Swedish rivals Volvo Trucks and Scania demonstrated automated trucks working in a mining’s fields. Volvo Trucks showed another case of cab-less truck in semi-closed logistics environment “Vera”.
Even-though automated or autonomous vehicles perceived as driverless cars, the truth is automation has several levels, and it will take decent time to become mainstream. Nowadays there are many up-running vehicles with so called ADAS, Advanced Driver Assistance Systems. No doubt ADAS makes driving easier, however the main concern of all ADAS’s is to increase road safety, and to avoid accidents.
ADAS has number of functions, some ADAS systems give Warning to driver like LDW “Lane Departure Warning”, CDW “Collision Detection Warning”, BSW “Blind Spot Warning”, PA “Parking Assist”,while some ADAS has Informative job like HUD “Head Up Display, SR “Sign Recognition”, and other ADAS actively take corrective actions like ACC “Adaptive Cruise Control”, LKA “Lane Keep Assist”, EMB “Emergency Mitigation Brake”, AWC “Automatic Wiper Control”.
ADAS mainly consists of sensor, processor, and actuator. Each system may require one sensor or more of same type or mix of sensors. Widely used sensors in ADAS are Ultrasonic, Cameras, Radars, and Li-dars. processor is keep monitoring sensors reading and compare it with set of reference values, to give control signals to Actuators. As long as ADAS is safety related, mainly it control over Brakes and steering.
SAE “Society of Automotive Engineers” defined 6 levels of automation from Level 0 “ No Automation” all the way to Level 5 “Full Automation”.
SAE differentiate automation levels on main 2 domains, Driver Responsibility/Awareness, and Nature of system either support or substitute driver. Upto Level 2 “Low Automation” driver is fully responsible about vehicle, and he have to maintain “Hands on - Eyes on”. Level 3 “Partial Automation” is unique as systems should be able to drive the vehicle, however driver should be ready to take over whenever System requested .
Currently due to the complexity of some ADAS systems, cost is still high, therefore we can find Level 3 in the Hi-End Vehicles. It is obvious there is no mature Level 4 and 5 product out in the market, and R&D and conditional testing being carried out.
Automation challenges arise since vehicles being operated in an extremely dynamic environment, ADAS should be able to carry out DDT “Dynamic Driving Tasks” in a sustainable manner regardless:
In the upcoming section, we will evaluate 4 fundamental challenges facing Autonomous Driving
Technical/Technological challenges
To reach Hi Automation level4 / Full Automation level5, all driving subsystems and components should be in operation all the time in a way to overcome the limitation of sensors, processor, storage, and actuators. In order to have a full control of all vehicle, ADAS should not be in discrete packaging , One super processor should be looking after all DDTs.
Any hardware has limitation due to many reasons, a list of challenges will be evaluated:
any sensor has a range of service for instance radar ranging from 80 to 200m length, Ultrasonic sensors maybe useful during parking, however proximity precision can be an issue in narrow spaces. Vehicle speed is varying all the time, sensors should be able to refresh its resolution in order to help processor taking the proper decisions at the right fraction of a second
Cameras angle limitation around 120 degrees, hence most popular 4 camera system is not enough to cover 360 of vehicle surrounding. Camera can get easily disturbed by direct light, shadows and reflection
Dense traffic of vehicles equipped with radars could be a challenge that radars will not be able to measure distance from reflection properly.
领英推荐
Latest ADAS systems is expensive as it not mass produced and manufacturing technology is new, i.e Li-DAR senor is the most expensive, cost of technology sometime play a role to slow down usage.
System should be equipped with adequate amount of onboard storage to save cameras hi-res video footages for Offline AI engine’s machine learning. Storage has its own limitation, and data should be transferred at ultra fast speed with no buffer lag to AI Engine. OEMs should decide on stored data retention period, and if it can be accessible to driver.
AI has been utilized for camera systems, a 360 models of vehicle and surrounding environment images being generated precisely, however AI applications with Radars and Li-DARs still under development.
Testing / Validation
Testing a prototype of AV is a complex task, and very costly at the same time. Many inputs which constantly varying affect the performance and response. There is no mature Level 4 or 5 product in the market right now, even though if you visit websites of main market players you can find millions of kms of safety driving in addition to millions of successful simulated driving.
Closed circuit testing is being used, however some OEMs built their own simulated cities and design different scenarios of testing to put their prototype in an environment can represent the actual driving situation. Some of the challenges facing AV testing will be evaluated.
Even though we see successful AV tests in closed environments like campus, however it seems like OEMs need more time to simulate more and much sudden actions in a regular driving day. It has been noticed either closed circuits or predefined route testing are not equal to the complexity of dense traffic cities.
To build a scaled city or urban for physical testing is very expensive and time consuming, which decelerate testing activities, delay product matuerity.
Lab simulations, and physical testing video footage should be utilized to the max and get benefited from the power of AI machine learning. Using AI could build more robust and complex testing scenarios, and could enhance final product reliability towards sever road conditions.
Complexity of testing, will make Automated public shuttling more appealing, and faster to deploy especially if it works in closed environment or defined route. In parallel commercial vehicles Platooning could be a faster deployment in dedicated trucks lane.
Legal Impact
Questions of Today Is it a Man or Machine World ?! Who will take the lead?!
These two simple questions, have a very ambiguous, and debatable answers. Robot will be soon everywhere taking assignments and doing critical and dangerous tasks, Autonomous Vehicles are no difference.
There is no doubt safety and zero accident are top priorities of automation, however in a mixed world of standard and automated vehicles, many are expecting issues to happen and so many queries are raising, i.e:
Strong Legalization is required to protect Mankind from Machine mistakes
Workforce
In commercial vehicle industry, many see automation as thread of losing thousands of jobs, mainly with frequently repeated tasks in a logistic hub, ports and shuttles. We have to find a way to develop workforce capabilities and redesign their tasks in a sustainable way.
Conclusion
Automated driving will be carrying bunch of advantages, at the same time disadvantages should be reviewed carefully. Deployment of Hi/Full Automation level for commercial vehicles should be happening with many considerations: