In this iteration of the Gradient Marine Newsletter we delve into the emerging effects automation and artificial intelligence have on modern vehicles.
Artificial Intelligence (AI) and autonomy are ideas gaining in popularity over the previous years. These words are almost always thrown around synonymously, while at their core aiming to solve the same problem. Both ideas are intended to streamline human efficiency by leveraging technology. We will discover autonomy, and automation, is really just a stepping stone for AI.
Automation refers to the use of technology and predefined algorithms to perform tasks with minimal human intervention. It involves the automatic execution of actions based on preset conditions to increase overall efficiency of the task. Automation aims to reduce manual labor where possible and more importantly reduce potential errors induced by human actions. Automation can be seen all over, simple tasks like sorting messages use automation as well as complex industrial manufacturing.
AI refers to the simulation of data driven algorithms in machines. Artificial Intelligence enables machines to perform tasks requiring some level of cognitive ability. These tasks can range from problem solving, decision making, and even learning from past behavior. In fact, AI will not work unless some sort of training data is present. The scope of functionality is much wider for AI in that it operates beyond what autonomy is limited by. Nevertheless, autonomy is better suited for supply chain tasks and AI excels at image generation. It is worth pointing out that AI is not a fully superior technology to autonomy, but excels at more tasks then autonomy. AI can utilize machine learning to analyze sources of data, generate a response, and then iterate the response thousands of times until a desired threshold. While AI driven tasks start out less refined than automation, the end product can be much more developed. The boundary conditions given to AI can be rough around the edges because it has the unique ability to learn on the job.
Similar to humans, a computer can operate in a dynamic environment and ultimately make an informed decision on what to do next. The idea of a dynamic environment becomes increasingly curious when we start to think about computer driven vehicles. Self-Driving cars for example, are a prime example of such an environment. In theory, a computer leveraging AI can receive data on other cars, people, and cyclists sharing the same stretch of road, much like a human processes the visual information and safely navigates the car to the destination.
Where can it go?
In an article written by TWI Global, the six levels of autonomous vehicles are articulated as the following:
- Level 0: The vehicle is 100% operated by the human driver and has no computer assists.
- Level 1: The vehicle has an installed driver assist system to help with steering, accelerating, or braking.
- Level 2: The installed driver assist system oversees most of the steering, accelerating, and braking inputs. However, the driver is still present and is required to pay attention to their surroundings, but performs less of the driving inputs.
- Level 3: The driving system is more advanced and the computer can drive the vehicle in specific conditions. In the remaining conditions, the human driver may be required to take over for safety reasons.
- Level 4: At level 4, the driving system can independently operate the vehicle in most conditions. The driver remains present to take over in unsafe conditions.
- Level 5: The vehicle is fully autonomous.
SAE International also has a nice graphic illustrating the levels.
Where is it now?
Most automakers have confidently achieved level 2 vehicles. The features seem simple but are quite game changing when experienced. Most drivers are familiar with basic cruise control; a mechanical system in which a secondary throttle cable is engaged and held at a specific location. This allows the car to maintain a regulated speed set by the driver. The system can be disengaged by applying pressure to the brake pedal. Cruise control is convenient for longer drives where the vehicle does not change speeds regularly. Autonomy is not only helpful during driving but also during parking encounters. Most cars on the road today will probably have some sort of reverse camera, with fancy versions where lines tracking the vehicle’s trajectory are displayed to the screen. Older cars may have sensors on the bumper that indicate the car to beep at an increasing rate when the sensor nears an object. These types of driver’s aid would be considered level 1 autonomy.
The level 2 version of cruise elevates this action. These cars also come with a traditional cruise control option, but many have an upgraded version as well. Additionally, the driver has the option to set a distance to the car ahead and maintain that distance. In essence, if the driver wants to stay 60 feet behind the car in front. The computer will adjust the brakes and accelerator to remain 60 feet behind through traffic. These cars also have lane assist. The car will either alert the driver they have swayed to the limit of their lane, adjust the wheel to keep the vehicle within the lines, or both.
Advantages of higher levels?
With a level 5 driving system developed capable of matching an alert human driving
performance, pedestrian safety is expected to improve. Barring any wide scale hacking, there are no obvious repercussions again in terms of safety. Last second lane changes, speeding, and running red lights will decrease as drivers migrate into level 5 capable vehicles. Fatigue would be eliminated from long distance drives. Autonomously driven vehicles could be a huge development in the trucking industry. The human driver could operate the delivery truck in and out of the high density city areas. Then, once on open roads, the computer would take over and let the driver rest before arriving at the next more populated area.
Autonomously driving cars would be an aid for those unable to drive. Old age and physical impairments are a few reasons individuals are prohibited from driving. Self-driving cars would eliminate the need for a family member or care taker to provide transportation assistance. Additionally, drivers under the influence would have a safe way to get home.
AI in the Air?
AI allows for the vehicles to go beyond human capability. A very recent application of AI has been for use in fighter jets. Air Force Lt. Col. Ryan “Hal” Hefron, the Defense Advanced Research Projects Agency (DARPA) program manager for Air Combat Evolution (ACE) had this to say about the first AI controlled fighter jet tested in December of 2022:
“We conducted multiple sorties [takeoffs and landings] with numerous test points performed on each sortie to test the algorithms under varying starting conditions, against various simulated adversaries, and with simulated weapons capabilities. We didn’t run into any major issues but did encounter some differences compared to simulation-based results, which is to be expected when transitioning from virtual to live. This highlights the importance of not only flight testing advanced autonomous capabilities but doing so on testbeds like VISTA, which allowed us to rapidly learn lessons and iterate at a much faster rate than with other air vehicles.”
The US government tested the first AI flown fighter jet. The X-62A VISTA( Variable Stability In-Flight Test Aircraft) is a modified F-16 fighter jet fitted with AI technology and sensors. The fighter was able to complete a 17 hour flight test including advanced fighter maneuvers and beyond-visual-range engagements. The test flight involved 12 different tests that could change throughout the test. The Air Force does have full sized remotely operated drones such as the MQ-1B Predator. The Predator drone was designed for a supportive role to aid in military operations. They carry a smaller payload than fighter jets and Due to the properties of AI, the team could land the X-62A, change the way a mission is programmed, and take off quickly.
The development of AI fighter pilots first started in 2020 when DARPA ran a convention for simulated dogfights. The idea was to understand how AI would perform in a dog fight. The AI-driven simulated fighter jet outperformed a human pilot in a first to five competition. The human pilot used a traditional joystick setup with virtual reality goggles, the human pilot did not win a single round. The AI pilot was able to track the human pilot and lock on with weapon systems with greater success.
From an article by Aerospace America
“The AI fighter scored simulated kills against the human every time, in part because the AI-driven fighter could aim its canon with superhuman accuracy from seemingly impossible attack angles, allowing the AI fighter to outmatch his human adversary in an old-fashioned, close-in, turning dogfight.”
Many of the drones used nowadays can already land and take off themselves. The autonomously operated drones do a great job in a well-known environment at easily distinguishable and pre-programmed actions.
In the future having a fleet of AI pilots would be a huge asset to the US Air Force. AI pilots would be used to help train human pilots. The skill level could be easily changed to give the human pilot the best practice. Additionally, human pilots could be slowly phased out. Removing the possibility of loss of human life as is present in a human flown mission. Nowadays, many US Air Force pilots run missions as escorts for bombers, a task well suited to near-future AI technologies. This would save these highly skilled personnel for missions and irregular operations or missions requiring a human pilot.
Currently the latest fighter jets performance is hindered by the human body. Often, the plane can withstand much more g-force than its human pilot. Most people can withstand 4-6Gs. A “G” refers to gravitational acceleration. Everyone currently standing on earth is experiencing 1G. Any additional gravitational force is a multiple. The F-22, a prominent fighter jet, can withstand 9Gs for the entirety of its gas tank. Fighter pilots only experience 9Gs for mere seconds due to the danger to the human anatomy, even holding a 6G pull can be fatal.
- TWI -Global Automobile automation levels: https://www.twi-global.com/technical-knowledge/faqs/what-is-an-autonomous-vehicle
- SAE Automobile automation levels: https://www.sae.org/blog/sae-j3016-update
- AI flown fighter jet: https://arstechnica.com/cars/2023/02/the-us-air-force-successfully-tested-this-ai-controlled-jet-fighter/
- MQ-1B Predator: https://www.af.mil/About-Us/Fact-Sheets/Display/Article/104469/mq-1b-predator/
- Dog Fight Simulation: https://www.darpa.mil/news-events/2020-08-26
- Introduction to AI pilots: https://aerospaceamerica.aiaa.org/departments/analysis-rise-of-the-ai-fighter-pilots/
- Future of AI pilots: https://www.airforce-technology.com/comment/ai-pilots-the-future-of-aerial-warfare