CSC477 Introduction to Mobile Robotics

Autonomous Navigation For Mobile Robots

Navigation can be defined as the ability of anyone to determine the location and then take to that place. For anyone to navigate the basic requirements are that they should have a GPS and second thing is to have a map to determine the path to reach to the specific goal. Now, when we are doing autonomous navigation, the same process is followed but over here there is no human in between. This is the type of process when we are asking a vehicle to determine the goal using a set of sensors which help the vehicle to move into the environment To a Desired Goal. Example a car, an aeroplane or a space craft journeying across the solar system or any submarine exploring the depth of the ocean. This is the process where the vehicle has the ability to make decisions and then act on its own. There are different levels of autonomy. First one is tele - operation where the vehicle is driven from a remote location and the vehicle is partially autonomous where it can control the vehicle such as stop it from hitting a wall. The second one is a fully autonomous vehicle where there will be no human interaction. We will focus on a fully autonomous vehicle.

Now, a fully autonomous approach can also be further divided into two different approaches. The first one is heuristic approach where the autonomy is accomplished through practical rules and behaviour. This will not guarantee an optimal solution. But this will achieve any goal immediately. This process has an advantage that it does not require complete information about the environment. The second approach is the optimal approach. This will require a lot more knowledge about the given environment. Also the plan and the resulting action will come from the maximisation of an objective function.

An example of a given heuristic approach will be the maze solving vehicle. Where the given rule is to drive forward and the wall should be at the left. This type of vehicle will proceed up and down the hallways until it reaches the goal. Here the vehicle does not have to map. This will not follow an optimal path but it works. Now, in the optimal approach, the vehicle will first build a model of the environment. And then it will figure out an optimal path to reach the goal. For example - autonomous driving. It makes more sense to give the vehicle to model the dynamic environment. Even in the case if the model is not perfect but to provide an optimal solution.

There can also be a case where the solution is not 100 percent heuristic and not 100 percent optimal. Both the approaches can be used to achieve the goal. For example when a car have to change the lane while moving on a road where there are fast moving as well as slow moving cars. Another example of autonomous vehicles is present in Amazon warehouses which are ground vehicles. This helps in moving packages while not hitting other vehicles. Another example is the vehicles that search and rescue in the disaster area.

Kalman Filter :

The Kaman filter is one of the important element for furious measurements. This is done using several sensors which will help in estimating in real time the state of a given robotic system such as a self driving car. The Kalman filter works in two stages : prediction and the correction. The Kalman filter was used by NASA in Apollo mission. The filter there helped the spacecraft through the lunar orbit. The engineers at NASA adopted the linear theory and extended that to non linear model.

The common linear filter is same as the linear recursive least square filter. The recursive least square updates the values of static parameters. The common filter is able to update an estimate of an evolving state. The goal of a common filter is to take a probabilistic approach of the state and then update that in real time while using two steps : these are prediction and correction.

When we consider the 1 D position of a vehicle and when we start at the initial probabilistic estimate at the given time k-1 , our goal will be here to use a motion model which will be derived from odometry. This will help in predicting the new state. The next step is to use the observation model example a GPS which will help to correct the position for time k. All the components the initial estimate, the predicted state and also the final corrected state all will be considered as the random variable. They can be specified by there means and the co variances. This is the way to take common filter as a technique which will help to fuse the information from the sensors. This will help in producing the final estimate of some unknown state. All the other factors are also considered such as uncertainty, motion and all the measurement.

The best thing about getting paper help from us is that you will get proofreading services by default. You don’t have to pay extra charges for the proofreading services if you seek assignment writing services from us. The proofreaders ensure that your paper is devoid of any mistakes. From the sentence structures and punctuation to relevance and context, our team of ghostwriters take care of everything. Expect nothing but higher grades when you opt for our assignment writing service. Do not hesitate to avail yourself of our essay writing service if essay writing isn’t your cup of tea. We have the best team to assist you.