College of Engineering, Pune

About: College of Engineering, Pune is a based out in . It is known for research contribution in the topics: Computer science & Sliding mode control. The organization has 4264 authors who have published 3492 publications receiving 19371 citations. The organization is also known as: COEP.

Implementation and analytical modelling of modified optimised link state routing protocol for network lifetime improvement

Abstract: Self-organising ad hoc networks need development of efficient routing protocols in terms of reliable routing and energy conservation. For dense ad hoc networks, optimised link state routing (OLSR) protocol is suitable owing to its multi-point relaying (MPR) feature. The authors have tried to make OLSR energy efficient by making effective neighbour selection based on residual battery energy of a node and traffic conditions that influence the drain rate of the node in the network. The authors have considered the multipath and source routing concept for route selection and a route recovery technique to tackle mobility issue efficiently. With these modifications the protocol becomes energy efficient and at the same time achieves balancing of network load. Simulation results of OLSR and the modified protocol OLSRM, show improvement in ‘Number of nodes alive’ against variation in pause time, speed and node density. In this study, the authors have tried to model the network behaviour using R software. The work also describes analytical modelling of the protocol performance in terms of energy aware metric, ‘Number of nodes alive’ with respect to simulation time as a common frame of reference. Using analytical modelling approach the authors can predict correct network behaviour for future instances.

Real time obstacle detection for mobile robot navigation using stereo vision

Abstract: Obstacle detection and avoidance is playing very important role in field of mobile robot navigation, space exploration and automation industries for safety of robot. In this study paper we are proposing a real time obstacle detection and avoidance algorithm using passive stereoscopic Kinect camera. The basic idea behind the obstacle detection method is to find a depth map of image captured by Kinect camera and map it with the real world coordinates. Camera can be used as non contact type of sensor for detection of obstacles unlike the classical sensors. The obstacle detection and avoidance is carrying out in both static and dynamic environment i.e. obstacles in environment can be stationary or moving. The proposed system is tested in indoor environment with raspberry pi 2 model with 640*480 pixels image size and with a frame rate of 30 fps. The system is simple, robust and efficient. By use of camera system as sensor the need for complex sensor arrangement can be avoided. By the experimental results we can conclude that the proposed system is robust and convenient.

Aminolysis of Poly(Ethylene Terephthalate) Waste for Recovery of Value Added Comonomeric Product

Abstract: Reaction of poly(ethylene terephthalate) waste (PETW) powder with hydrazine monohydrate (HMH) using 0.003 mol lead acetate as a catalyst was carried out in a batch reactor at 339K and at atmospheric pressure. Reactions were undertaken with various particle size ranges from 50–512.5 μm, and reaction time from 30–240 min at the interval of 30 min. Terephthalohydrazide (TPHD) and ethylene glycol (EG) were recovered during aminolysis of PETW. To increase the PETW conversion rate, external catalyst (lead acetate) was introduced during the reaction. The reaction product (TPHD) was deposited on the surface of unreacted PETW that was removed from the surface by introducing dimethyl sulfoxide (DMSO). Moreover, to improve the reaction rate cyclohexylamine (CHA) was introduced during reaction. Aminolysis of PETW was proportional to the reaction time. Depolymerization of PETW was inversely proportional to the particle size of PETW. Analyses of value-added products (TPHD and EG) as well as PETW were undertake...

A fast and light classifier for data streams

Abstract: Analysis of data streams is becoming a key area of data mining research, as the number of applications demanding such processing increases. Modern information technology allows information to be collected at a far greater rate than ever before. Machine learning offers promise of a solution, but the field mainly focuses on achieving high accuracy when data supply is limited. While this has created sophisticated classification algorithms, many do not cope with increasing data set size. When the data set size gets to a point where it could be considered to represent a continuous supply or data stream then incremental classification algorithms are required. When tackling with non-stationary concepts, ensemble of classifiers has several advantages over single classifier methods. They are easy to scale and parallelize, they can adapt to change quickly by pruning underperforming parts of the ensemble and they therefore usually generate more accurate concept descriptions and efficient results. But the effectiveness of an algorithm cannot simply be assessed by accuracy alone. Consideration needs to be given to the memory available to the algorithm and the speed at which data is processed in terms of both the time taken to predict the class of a new data sample and the time taken to include this sample in an incrementally updated classification model. This paper proposes a fast and light classifier for data stream classification.