Complex self-driving behaviors emerging from affordance competition in layered control architectures

Resolving uncertainty on the fly: modeling adaptive driving behavior as active inference

“…Our model also has similarities with the affordance competition hypothesis (Cisek, 2007) and its implementation in the computational driver behavior model by da Lio et al (2023). Specifically, both models are based on the continuous selection of action representations biased by value.…”

“…Adaptive driving behavior has been extensively studied for the past sixty years and there exists an influential tradition of (mostly) conceptual models offering explanations for the key underlying mechanisms (e.g., Senders, 1967;Näätänen and Summala, 1976;Summala, 1988;Fuller, 2005), which has been followed by more recent computational models addressing different adaptive driving behavior phenomena (Johnson et al, 2014;Pekkanen et al, 2018;Kolekar et al, 2020;da Lio et al, 2023).…”

“…A common denominator in most of these existing models is that adaptive driving behavior is, on the one hand, driven by a motivation to achieve goals and, on the other, by inhibitory motives such as the need to control uncertainty. The computational models reviewed above typically represent specific aspects of this phenomenon (e.g., visual sampling and time-sharing), except for a few notable developments of models with a more general scope (Kolekar et al, 2020 ; da Lio et al, 2023 ). However, a generic computational model of adaptive driving behavior, applicable across all types of scenarios and behaviors, is still lacking.…”

“…Another example of a generic computational model of adaptive behavior, related in some ways to Kolekar et al's (2020) field model, is the model developed by da Lio et al (2023) based on the affordance competition hypothesis originiating in neuroscience (Cisek, 2007). The concept of affordances were introduced by Gibson (2014), based on the earlier field of safe travel model (Gibson and Crooks, 1938).…”