Motor control in humans

Motor control in humans can be defined roughly as the way our brain controls our body. It is the way our Central nervous system (CNS) controls our bodily movements. Suppose you want to pick up an object. How do you do it and how does your brain convert the plan of picking up the object into reality by giving appropriate signals to the muscles. It turns out that even small things which we do so easily in life like picking up of an object or typing on a computer requires a series of complex processes which are controlled by the central nervous system (CNS).
The Motor control problem
Now exactly how the CNS does controls the actions, let's try to investigate it. Suppose if you want to pick up an object, what do you do? First you see the object/smell it, and then move your hand close to it, use sense of touch to determine the object's location and pick it up. First the brain gives a signal to the muscles and then the corresponding muscles relax or contract to produce the desired movement.

Parts of the problem

1. Firstly what happens in the brain, how does CNS plan the action and generate signals for the muscles.
2. Secondly how does the signal travel from CNS down to the muscles and
3. Thirdly how the signal to the muscle help in contracting or expanding it does. Let’s start with trying to investigate what happens in the CNS.
To plan a movement, the central nervous system (CNS) must transfer the sensory inputs into motor goals such as the direction, amplitude, and velocity of the intended movement. Then, to execute movements, the CNS must convert these desired goals into signals controlling the muscles that are active during the execution of even the simplest kind of trajectory of the hand.
The muscles contain of many muscular fibres in a bundle (See image), and it is established that each muscular fibre has a connection to a neuron axon ( See image of neuron) and that is how the signal travels from CNS to the muscle fibre. Thus there is a neuromuscular junction in each muscle fibre ( See image). A neuromuscular junction ( NMJ ) is the synapse or junction of the axon terminal of a motoneuron with the motor end plate , the highly-excitable region of muscle fiber plasma membrane responsible for initiation of action potentials across the muscle's surface, ultimately causing the muscle to contract. In vertebrates, the signal passes through the neuromuscular junction via the neurotransmitter acetylcholine.

On the basis of research it was found out that the CNS gives instructions to the spinal cord which has mechanisms for movement of muscles.
The problem CNS need to solve
It gets some inputs from the sensory organs about the object such as direction, velocity, etc and it has to use this information to generate a large number of signals to many muscles. Now this becomes an ill-posed problem in the sense that the solution may either not exist or may not be unique.Exaclty how the CNS solves this problem is a subject of recent studies.

The current view on the formation of arm trajectories is that the CNS formulates the appropriate command for the desired trajectory on the basis of knowledge about the initial arm position and the target's location. Recent psychophysical evidence supports the hypothesis that the planning of limbs' movements constitutes an early and separate stage of information processing. According to this view, during planning the brain is mainly concerned with establishing movement kinematics, a sequence of positions that the hand is expected to occupy at different times within the extrapersonal space. Later, during execution, the dynamics of the musculoskeletal system are controlled in such a way as to enforce the plan of movement within different environmental conditions.

Complexity of the problem
There are many computational complexities involved in the production of muscle forces. A variety of proposals have been made to explain these complexities. In theory, in a multijoint limb, the problem of generation forces may be addressed only after the trajectory of the joint angles has been derived from the trajectory of the endpoint—that is, after an inverse kinematics problem has been solved. Investigations in robot control in the late 1970s and early 1980s have shown that both the inverse kinematic and inverse dynamic problems may be efficiently implemented in a digital computer for many robot geometries. On the basis of these studies, investigators have argued that the brain may be carrying out inverse kinematic and dynamic computations when moving the hand in a purposeful way.

One way to compute inverse dynamics is based on carrying out explicitly the algebraic operations after representing variables such as positions, velocity acceleration, torque, and inertia. This hypothesis, however, is not useful because there is no allowance for the inevitable mechanical vagaries associated with any interaction with the environment.

Alternative proposals have been made that do not depend on the solution of the complicated inverse-dynamic problem . Specifically, it has been proposed that the CNS may transform the desired hand motion into a series of equilibrium positions (Bizzi et al. 1984). The forces needed to track the equilibrium trajectory result from the intrinsic elastic properties of the muscles (Feldman 1974).

One possible solution
To solve this problem we can make an analogy of it with computational motor control. It has a view towards parallels between computational modelling and theories in artificial intelligence and robotics, in particular robotics with anthropomorphic or humanoid robots. It can be structured according to the control diagram in Figure 1, which is commonly used in robotics and can also function as an abstract guideline for research in biological motor control.

File:Diagram 1.jpg

Figure 1 


Five major stages of motor control
This diagram distinguishes between five major stages of motor control: 

1. The higher level processing and decision making, which defines the intent of the motor system.

2. The motor planning stage: Planning the to be used by the CNS to know the method to pick up object.

3. Coordinate transformations: How does CNS know the coordinates of object to be picked up relative to its own coordinates?

4. The final conversion of plans to motor commands or instructions to be sent to spinal cord.

5. The pre-processing of sensory information such that it is suitable for control: This includes filtering out unnecessary information not required for the purpose.

Of course, the separation of the stages in Figure 1 might not be present in some control algorithms and in biological systems, but, as will be seen below, a conceptual differentiation of these stages will be useful for our discussion.

This suggests that the mechanism is feedback based rather than open loop , and control system techniques can be applied here. By feedback I mean that you continuously update the input based on the position of the object.

The motor command generation stage in Figure 1, which is usually associated with some of the main motor area in the primate brain such as the primary motor cortices and the cerebellum. It is now relatively well established that the central nervous system makes use of the computational principle of internal models, which are mechanisms that can mimic the input?output characteristics of the motor apparatus (forward models), or their inverse (inverse models) .Research in this area has started to focus on how multiple tasks are controlled with internal models, for example, objects with different weight or inertial properties. This topic is discussed in the literature as multiple model learning, contextual model switching, or mixture models.

Motor planning stage

The brain can be thought as a stochastic optimal controller. Noise in motor control can, in theory, arise from sensory (i.e. target localization), planning, execution, or muscular origins .Because noise, as sketched in Figure 1, is predominant in the nervous system, it is bound to affect motor control. What strategies does the central nervous system use to minimize the effects of noise on movements?

Van Beers et al showed that the variability observed in hand position after reaching movements is not explained by sensory or planning noise, but rather by noise in movement execution.Todorov and Jordan suggested a complete theory of stochastic optimal feedback control, which addresses motor planning, motor execution and redundancy resolution and that can account for a large body of experimental data. Note, however, that the computational complexity of learning stochastic optimal control is still daunting in nonlinear motor systems, even in theory.
How does motor learning occur

Many theories are proposed for how exactly the CNS learns to make body perform actions from intentions. Some of them are:

1. Error based learning.

2. Unsupervised learning: Methods such as reinforcement learning in which says that a system takes decisions to maximize the long term reward.

3. Also some people think that the mechanism may be based on Bayesian learning i.e. learning occurs by updating prior experience based on knowledge. Because Bayesian decision theory enables optimal integration of prior knowledge and sensed noisy information from multiple sources, the concept of Bayesian inference is particularly useful when uncertainty about variables needs to be incorporated into decision-making. A remarkable result of modern statistical learning theory is that many artificial and biological neural networks can be interpreted as Bayes optimal signal processing systems, despite possessing neither explicit knowledge of Bayes rule nor knowledge of the probability distributions of the involved variables.

Two interesting trends in motor control research can be seen recently. First, there has been a growing acceptance of complex computational models of brain information processing. Second, there have been several model-based experiments, in which complex models function as a guide to the experiment design and subsequent data analysis.

Main Menu

See Also

Main Menu