Comment on SAIC-GM RoboTest: It wants win-win, not win for itself

tech | 2024-08-30

On June 25th, SAIC General Motors unveiled their innovative breakthrough in the field of vehicle test certification through a live broadcast, showcasing their development of the "RoboTest unmanned intelligent vehicle testing platform."

Typically, automotive manufacturers either announce new car releases or new technologies. It is quite rare, if not unheard of, for a company like SAIC General Motors to announce testing technology in the field of vehicle development.

It is important to emphasize the following points:

1. The "RoboTest unmanned intelligent vehicle testing platform" is not aimed at consumers; rather, it is specifically designed for use by manufacturers.

2. While it can be simplistically understood as an unmanned driving system, it is distinct from the "intelligent driver assistance systems" currently available in the consumer market. RoboTest is an industrial-grade unmanned intelligent vehicle testing platform.As a veteran in the field of automotive R&D testing for many years, I believe that SAIC General Motors' recent showcase to the public is not merely to demonstrate their capabilities to ordinary consumers, but rather to "advertise" to other original equipment manufacturers (OEMs).

Advertisement

Other industry peers, please come to the Guangde Proving Ground for testing and make use of our latest testing robots.

At the same time, it also incidentally promotes SAIC General Motors' "high standards" in the field of automotive R&D, allowing consumers to recognize this quality-focused philosophy.

Why do I say this? To be frank, whether it's ordinary consumers or most media friends, they do not fully understand how many tests and verifications the OEMs conduct during the R&D phase, nor do they know exactly how these tests are carried out (which is also considered a corporate secret). It's quite good to have a general understanding.

Therefore, in this article, we will introduce and review SAIC General Motors' RoboTest from an engineer's perspective and discuss what impact it will have on the industry.01

What is the function of SAIC General Motors' "RoboTest"?

The "RoboTest" testing system by SAIC General Motors utilizes onboard robots to replace human test drivers for certain vehicle testing and certification tasks.

The entire system consists of two core components: the vehicle end and the cloud. The vehicle end controller integrates a driving robot system, positioning and perception equipment, and control modules, all of which are designed to be lightweight, standardized, and modular.

A crucial aspect is that it can be quickly installed and removed from the vehicle without damaging the original structure of the car – this is significant for vehicle compatibility, as if a car could only correspond to one set of equipment, it would be challenging to implement on a large scale.The cloud control center will be responsible for the remote configuration of test specifications and test paths, real-time monitoring and management, as well as the wireless transmission and analysis of test data, acting as a backend service.

In the promotional video, we can see that this system should have cameras and positioning systems installed on the top of the vehicle (two circular objects), which serve as the eyes of the test platform, seemingly for recognizing the route and positioning.

The other main components are the vehicle control mechanisms, including steering wheel operation, throttle operation, braking operation, gear shifting operation, and clutch operation mechanisms, used to directly control the vehicle. Although manual transmission models are becoming fewer in the current passenger car sector, they are not entirely absent, so integrating manual transmission machine operation is also necessary.

Moreover, from the backend monitoring interface, it appears that everything should be based on preset routes and programs, and it has not yet reached the level of automatic recognition, somewhat reminiscent of the nature of a line-following robot (this is just a metaphor and does not mean it is the same solution).According to SAIC General Motors, currently, the RoboTest platform has been widely applied in various test benches and road test sites at the Pan Asia Automotive Technical Center. This includes endurance, noise, emission, thermodynamics, and performance bench tests, as well as road tests such as Belgian roads, handling stability roads (G0 level), and hazardous item tests.

02

What are the benefits of the "RoboTest" system?

In the vehicle testing and verification system, there are numerous verification items. This includes the series of tests mentioned by SAIC General Motors in the previous text, in addition to many objective and subjective test items.

Taking the vehicle durability and reliability test field and the chassis performance test field that I am personally familiar with as examples, the durability and reliability tests include not only endurance bench tests but also enhanced road tests at the test site, high-speed loop tests, public road tests, and three-high tests (high temperature, high humidity, and high altitude), among others.In the high-altitude, high-temperature, and high-humidity ("three highs") tests, there are various test items, including environmental adaptability tests, cold start tests, heating tests, defrosting tests, low-temperature charging, low-temperature soaking, low-temperature discharging, snow throwing tests, and heating and cooling system driving evaluations in the cold region tests, among others.

Also included are air conditioning cooling tests, power battery temperature rise tests, high-altitude braking tests, dust throwing tests, and high-temperature charging and discharging tests in the high-temperature plateau tests; while in the field of objective chassis performance testing, there are parking & idling steering tests, steady-state turning tests, step input tests, sweep frequency tests, sine steering input tests, pitch tests, slalom tests, emergency avoidance tests (the well-known moose test is a type of emergency avoidance test), and braking tests, etc.

If one were to elaborate, it could take days and nights to discuss.

▲ A durability test cycle at a certain test track

In the vehicle validation system, many of the above tests are open-loop standard tests that require controlling test variables to ensure consistent test conditions during the tests, thereby evaluating certain aspects of vehicle performance.Here is the translation of the provided text into English:

For instance, during the vehicle development process, accelerated road tests are conducted in test facilities according to standards set by the original equipment manufacturer (OEM). These are cyclical tests where the vehicle must complete a certain mileage following a predetermined route, road surface sequence, and speed. The roads in these tests are characterized by enhanced features, such as the well-known Belgian block road and manhole cover road, which subject the vehicle to relatively strong impacts.

Compared to regular roads, these enhanced roads have a higher amplification factor, allowing the test vehicle to complete the entire life cycle of durability verification in a shorter distance and time. This type of test is considered an accelerated test.

Another example is the objective handling stability test known as the step steer test. It requires the vehicle to maintain a straight-line speed of 100 km/h for at least 2 seconds with a fixed throttle, then quickly turn the steering wheel to an angle that corresponds to a lateral acceleration of 0.2g, 0.4g, or 0.6g when the vehicle reaches a steady state. This is done to measure the vehicle's roll and yaw response.

The shorter the response time, the better and more sensitive the vehicle's steering response is.

▲ Processing results of the step steer test for a certain vehicle model.Similar experiments, which have standard operating procedures, require the control of certain variables—this is actually most suitable for "robot" operation.

 

Because human operation is influenced by factors such as personal skill level, physical condition, and mindset, it is impossible to maintain identical operations each time. However, when using robots for operation, the success rate, accuracy, and consistency of the experiment are higher.

 

Additionally, there are many items in the tests conducted by the main plant that pose significant risks—such as the snow and dust tests I mentioned earlier in the high-altitude, high-temperature, and high-humidity tests.

 

During these tests, a test vehicle needs to follow another test vehicle, using the snow or dust raised by the front vehicle's travel for the experiment. This is to simulate the conditions of a user driving behind another vehicle in a dusty environment or after snowfall, in order to verify the vehicle's body seal performance, air filter filtration performance, and so on in a dusty environment.

 

The danger of these two tests lies in the fact that the dust and snow will enter places such as the front compartment, chassis, and body, causing abnormalities in the vehicle. Therefore, the risk of these two tests is that the following vehicle is almost driving blind behind the car, especially in the dust test, where visibility is almost zero.So, from this perspective, "RoboTest" is undoubtedly a good tool:

Firstly, it can handle repetitive and mechanical tasks, especially the ability to work "non-stop," without having to consider the rest needs of the testers.

Secondly, it can address some work conditions that are relatively dangerous for testers, ensuring their personal safety in testing projects that involve risks.

Additionally, from the promotional video, "RoboTest" appears to have "night vision" capabilities. Typically, if tests such as high-speed loop tests or accelerated endurance tests are conducted at the test site, tests in the latter half of the night would require special permission due to the higher safety risks associated with poor visibility at night. Fatigue can easily lead to safety incidents, and the rescue capabilities after an accident at night are not as effective as during regular daytime working hours. Therefore, tests are generally not conducted in the latter half of the night unless it is an emergency.

In this way, utilizing robotic testing can significantly reduce the risks associated with night-time testing, and the development and testing cycle can be shortened as much as possible, which is definitely a good thing for the current pace of research and development.Many people, upon seeing this, would find this system quite novel. However, it did not emerge out of thin air, as there have been many similar purposed auxiliary testing tools for a long time:

For instance, in stability steering tests, there are steering robots capable of conducting tests with fixed steering angles and speeds, ensuring test success rates and consistency that far exceed those of manual tests. Additionally, there are no shortage of devices that automatically operate the throttle and brake pedals in bench tests.

Nevertheless, previous auxiliary testing systems could indeed only perform certain single functions and did not possess the capability of "autonomous driving."

"RoboTest" belongs to the most advanced testing phase auxiliary tools currently available on the market, and it can fully be referred to as a "robot" to demonstrate its level of intelligence.Does the "RoboTest" system have any flaws?

From the introduction and video provided by SAIC General Motors, I summarize this system as an "unmanned driving test assistance system"—I call it "assistance" because I believe it currently cannot achieve "complete replacement of manual labor," as there are many tests it may not be able to perform for the time being.

As previously mentioned, for standard, quantitative, and repetitive test tasks, robots are indeed a good helper. However, in the current world, almost all automotive companies' verification systems cannot rely solely on this type of testing to fully verify the vehicle; a combination of subjective and objective tests is required to adequately verify the vehicle.

Even for durability and reliability tests on test tracks, which are standardized conditions, the goal during testing is not only to understand the vehicle's durability performance at the end of the test period but also to discover any issues with the vehicle during the test process, including chassis noises, interior noises, system failures, and poor functionality, etc.

I believe the "RoboTest" system by SAIC General Motors may not be able to achieve this, as these issues involve many experiential problems that require test drivers or engineers to experience and then identify. Moreover, from the official materials available, the "RoboTest" system currently does not seem to have the capability for fault diagnosis.Additionally, currently, all projects related to subjective evaluation tests, "RoboTest" cannot handle either, such as the tuning, acceptance, and evaluation of chassis performance. Even if a robot could conduct the entire set of objective chassis performance tests and automatically process data to draw conclusions, it still wouldn't be able to fully assess the quality of the chassis performance.

This is because, although there are many objective test items available today that can reflect some aspects of chassis performance, the final acceptance and tuning of the chassis are still primarily based on subjective evaluation. Objective data sometimes merely serves as a verification or supplement to the subjective assessment. This is where the importance of a "vehicle subjective evaluation engineer" lies. And since it is "subjective," it is "human subjectivity," which naturally cannot be done by a robot.

Chassis tuning is even more impossible. It is not feasible to test every set of solutions with a robot, as this would not solve the issue of efficiency, thus losing the advantage of the robot's high efficiency—this is also one of the reasons why many people say that chassis tuning is a dark art. From an engineer's perspective on the chassis: excellent objective data never guarantees a good subjective experience.

Apart from the above discussion on the scope of applicable test items, there is a flaw in "RoboTest" that I believe still needs continuous improvement.

That is, as observed from the videos, the robot's ability to maintain its path during various tests is similar to the "lane-keeping assist feature" of many vehicles today. It needs to constantly adjust its direction within the lane. This is quite evident from the high-speed loop test videos, where the vehicle can be seen swaying back and forth, and the same is true when driving on simulated urban roads.Automatic lane keeping, where robots strictly control the vehicle to stay in the center of the lane; however, in real-world environments, human test drivers have a certain margin of error and do not adjust the direction as frequently as robots. Over-adjustment may lead to premature damage to the steering system or tires, which could also have a certain impact on test results. It is believed that this issue will be properly optimized and resolved in the later stages.

04

The significance of "RoboTest" for the industry

Initially, I watched the endurance test of the GL8 Land尊PHEV and noticed that this test was conducted by "RoboTest". The actual range of 1401 kilometers made the test conducted by robots feel more authentic: as robots are not influenced by subjective or objective factors, there is no need for the enhancement of a "golden right foot", and the test on real roads is more realistic than the data obtained from laboratory tests.

Currently, the Guangde test track receives over 5000 vehicles from various domestic manufacturers for testing each year. Not long ago, Xiaomi SU7 emphasized the rigor and professionalism of its testing by specifically stating that it was tested in Guangde.

RoboTest, due to its high degree of vehicle adaptability, can basically test all types of vehicles, and since it is currently serving the Guangde tests, its emergence undoubtedly has very significant importance for establishing Guangde's leading position in China's automotive industry. Whether it is for advertising or highlighting industry status, this wave of RoboTest has highlighted the technical strength of SAIC General Motors, which is unquestionable.However, my personal impressions go beyond this.

The recent hype surrounding the term "rolling" ultimately did not translate into care for "people," which I personally find regrettable. Everyone is arguing over who is right and who is wrong, but whether it's layoffs or overtime work, the issue of "people" is seldom mentioned, and it's difficult to discuss how to effectively address it.

However, in the entire application scenario and value of RoboTest, I think we can all sense one thing: attention to the individual. This is not just something that SAIC General Motors claims; the endurance testing of the new car in Guangde, which exceeds 6 million kilometers, inherently embodies a concern for the safety of the users, a concern that is "human by nature."

At the same time, in terms of mechanical repetition and dangerous testing, replacing humans with robots is also a form of attention to engineers—this point, as you can see from our analysis above, the value of RoboTest is not to completely replace engineers. In fact, I believe that in the realm of subjective evaluation, the emergence of RoboTest actually highlights the more valuable and critical nature of the "human by nature" assessments made by engineers.RoboTest is an intelligent auxiliary tool that takes care of tedious and repetitive tasks, while decisive subjective evaluations are left to humans. The role and value of humans are, in fact, further highlighted.

I believe that the very manufacturers who can create all of this deserve respect. And at this moment, as we understand and contemplate these seemingly dry technologies, through these complex mechanisms, we can see another kind of original intention of the car makers.

What touches me the most is that it's not about winning on its own. It's about sharing with the entire industry to achieve a win-win situation.

05

Driving Summary

Robots are indeed a good helper in the field of testing. For quantitative tests such as fixed throttle, fixed speed, custom power, fixed price speed, fixed angle, and fixed rotation speed, they are nothing short of a "godsend," able to improve the precision and consistency of tests. They also greatly reduce the risk to testers in dangerous test projects. In the future, there will be no more worries about electric leakage when electric vehicles are tested for water immersion, nor will testers need to wear protective suits for such tests. These are all advancements in testing.However, from what is currently known, it is not as perfect as we might imagine, and there are some minor flaws. But I believe these flaws will definitely be resolved in subsequent iterations. It is inevitable for a new system to have some minor flaws; as long as the direction is found, solving them is just a matter of time. SAIC General Motors has provided direction for all fellow testers, which is also excellent.

Additionally, as mentioned earlier, I still believe that the system from SAIC General Motors is primarily aimed at automotive companies, offering a more diverse range of testing options for businesses. One can imagine the immense commercial value for SAIC General Motors if major manufacturers could have several sets of equipment at their disposal—perhaps, in the future, SAIC might become a world-renowned equipment supplier with this system.

In summary, this is not a technology that SAIC General Motors can use exclusively; as a supplier, they are willing to share this industrial technology with the entire industry. The human care reflected behind these technologies also shines in a different light in the current "dog-eat-dog" world.