How Do Humans Learn to Control Unstable Objects? Studies of Model-Based Planning and State-Dependant Force Control

[00:68 . 84] mussa - ivaldi虽然我们对协调手臂运动的机制的理解取得了重大进展，但对熟练操纵物体所涉及的感觉运动过程知之甚少。处理银器和刷牙等常见活动涉及控制可能不稳定的物体。“不稳定”一词指的是一个人行为上的小偏差可能会导致严重后果的情况。例如，手稍微倾斜一下就会把勺子里的东西洒出来。该研究将研究人类在简单不稳定物体上的学习和实验操作任务的表现，并通过机器人模拟来实现。我们的研究旨在更好地理解在处理不稳定物体时反馈机制和预编程行为之间的相互作用。这项工作围绕三个具体目标展开。首先，我们提出的实验将测试在物体操作过程中是否形成内部模型，同时保持不稳定物体的平衡。早期对手臂控制的研究表明，中枢神经系统利用肢体动力学的内部模型来推导神经肌肉指令，以完成成功的伸展运动，但尚不清楚类似的策略是否用于物体操纵。其次，当我们处理一个物体时，我们既接收视觉信息，也通过手与物体之间的机械相互作用传递信息。这些信息对于产生适当的纠正措施是必不可少的，但在学习过程中可能不会以同样的方式使用。所提出的实验将验证这样的假设，即在熟练掌握操作任务后，受试者不再依赖于对物体的视觉监控，而更多地依赖于肌肉和反射力学产生的自动纠正。在最后一组实验中，我们将寻求确定学习处理不稳定物体的有效训练策略。我们期望这项工作将促进我们对健康人如何处理不稳定物体的理解。这一目标的实现将为中风和其他神经疾病导致的运动功能丧失的康复提供更丰富的知识基础。

9900684Mussa-IvaldiWhile significant progress has been made in our understanding of the mechanisms for executing coordinated arm movements, little is known yet about the sensory-motor processes involved in the skillful manipulation of objects. Common activities such as handling silverware and brushing teeth involve the control of potentially unstable objects. The term 'instability' refers to situations in which small deviations in one's actions can have radical consequences. For example, a small tilt of the hand can spill the contents of a spoon. The proposed research will study human learning and performance of experimental manipulation tasks based on simple unstable objects implemented by means of robotic simulation.Our investigations are aimed at better understanding the interplay between feedback mechanisms and preprogrammed actions in dealing with unstable objects. The work is organized around three specific aims. First, the proposed experiments will test whether internal models are formed during object manipulation, while maintaining the equilibrium of unstable objects. Earlier studies of arm control suggest that the central nervous system makes use of internal models of limb dynamics in order to derive the neuromuscular commands needed to carry out successful reaching movements, but it is not known whether similar strategies are used for object manipulation. Second, when handling an object we receive both visual information and information conveyed by the mechanical interaction between the hand and the object. This information is essential to generate the appropriate corrective actions but it may not be used in the same way as learning progresses. The proposed experiments will test the hypothesis that subjects become less reliant on visual monitoring of the object and more reliant on automatic corrections produced by muscle and reflex mechanics after attaining proficiency in a manipulation task. In a final set of experiments we will seek to determine effective training strategies for learning to handle unstable objects.We expect that this work will advance our understanding of how healthy people handle unstable objects. The achievement of this goal will provide a richer knowledge base for rehabilitation of motor functions lost to stroke and other neurological diseases.