Our patent-pending cogniBOT™ AI architecture realistically simulates how humans perceive, think, and act. They reproduce the entire sensorimotor chain of information processing including all limitations of human perception and behavior in traffic.
Human behavior emerges from a complex sequence of processing steps in the central nervous system. Many of these are highly relevant for traffic situations and are explicitly simulated by the cogniBOT™ architecture. The following examples illustrate how characteristics and limitations of human information processing influence behavior in traffic.
When driving a car, visual perception is the #1 sense for gathering relevant information. Important characteristics of human vision are its limited visual field and the non-uniform distribution of photoreceptors in the eyes, allowing for sharp vision only in a small central area of about 2°, the fovea. Humans compensate for these limitations by purposefully directing their gaze towards relevant objects or events. Nevertheless, in the context of road traffic, other road users can be overlooked or distances can be overestimated, resulting in critical situations.
Attention and Gaze Control
Eye movements are essential to orient our gaze towards important objects in our visual field. Where we direct our gaze is tightly coupled to our attention. On the one hand, salient stimuli attract our attention in a bottom-up manner. On the other hand, humans can volitionally orient their attention towards relevant objects or events.
Attention controls what we perceive and do both in top-down and bottom-up processes. In traffic, many different stimuli like traffic lights, road signs as well as other road users compete for our attention.
Humans continuously form an internal representation of their environment. However, the capacity of human working memory is limited. As a rule of thumb, the ‘magic number 7’ is commonly regarded as the amount of entries humans can retain in their working memory.
This limitation can lead to problems especially in complex traffic scenarios, in which it can be difficult to keep the overview of many simultaneous events.
Prediction & Decision Making
One advantage of human drivers in contrast to most autonomous vehicles is the ability to anticipate future states of the present traffic scenario. This capacity is based on current observations combined with prior knowledge about the causal structure of the world. Thus, humans can react quickly and efficiently even in complex traffic scenarios. Important decisions about what to do next, e.g. whether and when to overtake a slower car ahead, can be reached within milliseconds.
Emotions & Physiology
Human factors such as emotions, moods and physiological states shape human perception, cognition and action. Excitation increases the affinity for distractions, anger and aggressiveness can lead to risk-taking behaviour, tiredness can increase reaction times, etc. But also intoxication and the influence of medication have detrimental effects on driving performance.
All these influences mean that even the same person behaves differently in traffic depending on their emotional state. They are a main cause of human driving errors and accidents.
Perceptive and cognitive processing becomes effective and visible in the form of motor action. Motor capabilities are highly dependent on experience and age.
On the one hand, novice drivers need to learn an internal model of the vehicle, which links desired trajectories to actuator engagement such as pedal position and steering angle. On the other hand, with increasing age, motor execution becomes slower and motor ranges become increasingly restricted.
Real-world traffic is filled with all different kinds of road users. Thus, any autonomous vehicle needs to be able to cope with the resulting motion profiles.