Bid Data Paper

Assignment 4

Part 3:  

A Generic Robot Database and its Application in Fault Analysis and Performance Evaluation

The paper discusses the data collection in robotic systems and its usage for fault tolerance purposes. The data generated in such systems is very high. Usually that data is not stored. But if stored, it can be analyzed to get information about performance.

The data generation is not in a constant format. A system consisting of a NoSQL database is proposed. MongoDB stores data in form of documents, which consist of similar records. It is designed to be scalable and to support replication. These features are very useful in case of distributed systems consisting of multiple robots.

Data storage in such a system can be very useful as it can be used for learning when the robots are idle, also for storing a lot of information during the course of time.

The database system can be integrated with the middleware of the robotic system and then can be used to:

• Store data on a large scale
• Study fault tolerance
• Evaluate performance using map reduce

The versatile structure of MongoDB is useful in data storage. The structure of the data being inserted is not declared in a particular way. The insertion and retrieval of data is very fast in MongoDB.

In case of fault analysis, the fault may be due to a variety of causes. It can be faulty sensor, or processing or wrong data. Considering the huge amount of data generated, its important to scan all the incoming data for the source of errors.

Also the knowledge the robot uses is based on previous encounters. The data storage can contain lessons and infer wisdom form that for future uses.

Performance evaluation uses Map reduce to group the data into successful and unsuccessful records.  1 or 0 is emitted and the reduce function sums it up for an overview. The finalizer can be used for calculation of variance or average of the data.

Experiments are conducted across two scenarios. One consists of the data recorded is used to analysis of the fault. In the second one the data is used to obtain a quantitative analysis of the performance using map reduce.