Dual Process Control Models in the Brain and Machines with Application to Autonomous Vehicle Control

自动驾驶汽车控制（autonomous vehicle control，AVC）是一个快速发展的领域，它有望改善汽车的性能、燃油经济性、排放水平、舒适性和安全性，传统的控制方法在受限的环境下可以产生足够的结果，但设计和计算成本高，并且在真实的环境变化下很脆弱（风、其他车辆的接近度等）。因此，迫切需要AVC的替代方法。一种特别有前途的替代方法是将任务分解为一组子任务，每一个都在有限的条件范围内有效，并在需要时在它们之间切换。在过去的几年里，Hussain博士在斯特林的研究小组一直在为这种“模块化学习控制器”开发一种新的框架。从一组动作或行为中进行选择的问题也是动物的一个核心问题。越来越多的证据表明，所有脊椎动物都使用一组大脑中央核团--基底神经节--来帮助解决这个问题。在过去的十年里，格尼教授实验室的研究一直在开发基底神经节如何支持行为选择的计算模型。因此，我们相信基底神经节在所有脊椎动物的大脑中都扮演着中央“选择器”或“开关”的角色，因为它们检查对行为的请求，并且允许选择最紧急或最显著的请求用于行为表达。考虑到动物中AVC和动作选择这两个问题领域之间的相似性，该项目旨在利用心理学和神经生物学的新成果（在格尼教授的实验室发现）并将其应用于Hussain博士小组正在开发的AVC控制器。在这方面，动作选择的一个方面似乎特别有前途，这是两种普遍的行为选择模式。要理解这一点，请考虑以下情景。首先，想象一下早上起床后不久在自己的厨房里泡茶。你可能确切地知道该做什么，而不必有意识地意识到这一点--茶、牛奶、糖、水壶和水龙头的位置都是很好的学习，与这些位置的物体互动所需的运动动作也是如此。事后的反省使我们使用这样的术语：“我是在睡梦中做的”或“我是在自动驾驶”。现在，如果你是第一次住在朋友家，考虑做同样的任务。一个完全不同的策略似乎被使用。因此，我们必须警惕，探索，使用我们希望能很好概括的高级认知知识（例如，我们希望茶放在水槽旁边的橱柜里，而不是放在客厅里）这两种控制模式在心理学文献中分别被公认为自动和控制或执行处理。此外，斯特林开发的新AVC系统有两个主要组成部分：一个高级“监督”控制器和一套基本的（但适应性强）控制器，指导实际车辆行为。我们相信，与执行和自动控制的生物概念的相似之处高度表明神经科学和控制理论领域在这方面存在着相互富有成效的互动。因此，虽然我们的总体目标是利用控制工程系统和动物大脑之间的一系列相似性，我们将特别关注自动化和控制的概念，（或执行）处理以及它们如何映射到上述类型的模块化AVC解决方案。结果应该是新一代实时AVC控制器，更直接地受到生物学思想的启发。我们将与我们的工业合作伙伴（工业系统控制和SciSys）合作，评估这些新型控制器在常规道路驾驶和行星漫游车背景下的好处。

The field of autonomous vehicle control (AVC) is a rapidly growing one which promises improved performance, fuel economy, emission levels, comfort and safety.Application of conventional control methods can generate adequate results under restricted circumstances,but have high design and computational costs and are fragile under real environmental changes (winds, proximity of other vehicles etc).There is therefore a pressing need for alternative approaches to AVC.One particularly promising alternative is to break the task into a set of sub-tasks,each valid over a restricted range of conditions, and to switch between them when required.Dr Hussain's group in Stirling has been developing a novel framework for such'modular learning controllers'over the last few years.The problem of selecting from amongst a set of actions or behaviours is also a central problem for animals.There is growing evidence that a set of central brain nuclei -the basal ganglia- are used by all vertebrates to help solve this problem.Research in Prof Gurney's lab has,over the last decade,been developing computational models of how the basal ganglia support behavioural selection.Thus,we believe that the basal ganglia act as a central 'selector' or 'switch' in all vertebrate brains,in that they examine requests for behaviour and allow the most urgent or salient requests to be selected for behavioural expression Given the similarity between the two problems' domains of AVC and action selection in animals, this project aims to leverage new results from psychology and neurobiology (discovered in Prof Gurney's lab) and apply them to the AVC controllers under development in Dr Hussain's group.One aspect of action selection which appears particularly promising in this respect has to do with there being two general modes of behavioural selection.To see this,consider the following scenarios.First,imagine making tea soon after getting out of bed in the morning in your own kitchen.You probably know exactly what to do without having to consciously be aware of it--the location of the tea,milk,sugar,kettle and water-tap are all well learned, as is the motor actions required to interact with the objects in these locations.Introspection after the event leads us to use terms such as;`I did it in my sleep' or `I was on auto-pilot'.Now consider doing the same task if you are staying at a friend's house for the first time.A completely different strategy appears to be used.Thus,we have to be alert, explore, and use high level cognitive knowledge that we hope generalises well (for example,we hope the tea will be in a cupboard near the sink, not in the living room)These two modes of control are well recognised in the psychological literature as automatic and controlled or executive processing respectively.There is also growing neurobiological evidence for the existence of different control regimes, supported by different brain systems.In addition, the new AVC systems developed at Stirling have two major components:a high level 'supervisory' controller and a set of basic (but adaptable) controllers that direct the actual vehicle behaviour.We believe the similarities with the biological notions of executive and automatic control are highly indicative of a mutually fruitful interaction between neuroscientific and control theoretic domains in this regard.Thus, while our general aim is to exploit a range of similarities between systems in control engineering and the animal brain, we will focus specifically on the concepts of automatised and controlled (or executive) processing and how they might map onto modular AVC solutions of the kind described above.The outcome should be a new generation real-time AVC controller, more directly inspired by the biological ideas. We will work with our industrial partners (Industrial Systems Control and SciSys) to evaluate the benefits of these novel controllers within the context of regular road driving and planetary rover vehicles.