Alka Bhushan - Academia.edu (original) (raw)
Papers by Alka Bhushan
Indoor Evacuation Planning
Geospatial Infrastructure, Applications and Technologies: India Case Studies, 2018
Making efficient evacuation plans for large buildings is an important task in operating a large b... more Making efficient evacuation plans for large buildings is an important task in operating a large building infrastructure. The plans require a capacitated building network, well-defined exit points, and expected occupancy of a building to analyze evacuation time in different scenarios. Such type of analyses helps us in both the re-design to improve evacuation times as well as for creating more optimal allocations of the building spaces. In this chapter, we model the problems in the following two different types of situations and present heuristic algorithms which are extensions of the existing capacity-constrained route planner (CCRP) algorithm: (i) Additional exits with ladders need to be placed for faster evacuation, and (ii) all evacuees cannot be evacuated by themselves and require help from rescuers. For large and complex buildings, it is difficult to model indoor network and understand an evacuation plan for different scenarios without any user interactive interface. We have designed an evacuation planner which is publicly available and can be used for making indoor network from floor images, creating multiple scenarios, computing evacuation plans of these scenarios, and performing sensitivity analyses of an evacuation plan to evacuees’ behavior.
Mining Swarm Patterns in Sliding Windows over Moving Object Data Streams
Proceedings of the 25th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems, 2017
Several emerging applications such as traffic management and urban emergency response systems oft... more Several emerging applications such as traffic management and urban emergency response systems often need to identify groups from recent window of the moving object data. These requirements mandate algorithmic solutions that are time and memory efficient for adding new data incrementally and removing stale data. In this paper, we consider the problem of finding closed Swarms over a sliding window. The key challenges in computing closed Swarms over a sliding window are: (i) Search space for computing closed Swarms from new data is large (ii) Removal of old data leaves many non-closed Swarms which need to be identified and deleted. None of the existing methods are efficient in adding new data and removing old data for large datasets. This paper presents an efficient incremental graph based method for computing Swarms over sliding windows. We use a real dataset to show the performance of our method. The complexity analysis as well as experimental results demonstrate that our method is significantly faster than the existing incremental method over sliding windows with increased memory requirement. In particular, our method is shown to be 7-13 times faster with 3-5 times memory overhead in all the experiments.
A Distributed Learning Simulation Platform for Edge Hierarchies
2020 International Conference on COMmunication Systems & NETworkS (COMSNETS), 2020
We develop a distributed learning simulation platform that allows users to create multi-level Edg... more We develop a distributed learning simulation platform that allows users to create multi-level Edge hierarchy for a given application by simulating resource constrained Edge devices and communication links amongst them. The resulting Edge computing hierarchy is used to run a given DNN for the application in a data distributed fashion, rolling up learned parameter values up a hierarchy of parameter servers that merge parameters received from the lower levels. The root of this hierarchy has the latest model, which is then pushed lazily back down the tree to the Edge servers. The platform can be used to study cost vs accuracy analysis of a given application for different Edge hierarchy configurations. We use handwritten digit recognition problem as a case study to show the usefulness of our platform.
Mining Swarms from Moving Object Data Streams
Current methods for mining groups from moving object data work with entire data stream. However, ... more Current methods for mining groups from moving object data work with entire data stream. However, there are several emerging applications such as traffic management and urban emergency response systems which often need to identify groups from recent window of the moving object data. These requirements mandate algorithmic solutions that are time and memory efficient for adding new data incrementally and removing stale data. In this paper, we consider the problem of finding closed swarms over a sliding window. Large search space for computing closed swarms from the new data is the main key challenge in computing closed swarms over a sliding window. None of the existing methods are efficient for this. This paper presents an efficient incremental graph-based method for computing swarms over sliding windows. We demonstrate the performance of our method on two real datasets. The results show that our method is significantly faster than the existing incremental method over sliding windows w...
Proceedings of the 1st ACM SIGSPATIAL International Workshop on the Use of GIS in Emergency Management - EM-GIS '15, 2015
This paper focuses on indoor evacuation path planning problem where the objective is to find evac... more This paper focuses on indoor evacuation path planning problem where the objective is to find evacuation paths for each evacuee such that the evacuation egress time is minimized. Since paths are dependent on the distribution of evacuees, initial positions of evacuees are required to find optimal paths during emergency. Instrumenting a building to obtain initial positions and count of people in a building is very challenging (and costly), and hence evacuation plans are prepared for a few expected distributions. Generally, a standard plan based on a predominant distribution of evacuees is posted on the walls inside an indoor facility for people to follow. However, the actual distribution may be another distribution from among the possible distributions. In this work, we consider the problem of finding the distribution that prevails at the evacuation time so that evacuees can be guided to follow the optimal paths (leading to minimum time) rather than following the standard plan. We propose a cost-effective solution to this problem by observing movement of people within a specified time period, labeled distribution detection window, using minimum number of optimally located sensors. This is in contrast to existing approaches which assume that unlimited sensors are available to instantaneously obtain the exact distribution of evacuees at the time of incidence. To our knowledge this paper presents the first formal evacuation planning approach that enables the user to optimally tradeoff the delay in distribution detection with the cost of the deployed sensor network used to obtain this distribution. Our approach is based on the popular heuristic denoted as Capacity Constrained Routing Planner (CCRP). Our approach is illustrated by a set of experiments on two case studies. The results demonstrate that evacuation plans obtained using minimum number of sensors are better than the standard plans and are comparable to evacuation plans computed using unlimited number of sensors.
An I/O Efficient Algorithm for Minimum Spanning Trees
Lecture Notes in Computer Science, 2015
An O\mathrm{Sort}E\cdot \log \log _{E/V} B$$OSortEi¾?loglogE/VB I/Os algorithm for computing a ... more An O\mathrm{Sort}E\cdot \log \log _{E/V} B$$OSortEi¾?loglogE/VB I/Os algorithm for computing a minimum spanning tree of a graph G=V,E$$G=V,E is presented, where \mathrm{Sort}E=E/B\log _{M/B}E/B$$SortE=E/BlogM/BE/B, M is the main memory size, and B is the block size. This improves on the previous bound of O\mathrm{Sort}E \cdot \log \log VB/E$$OSortEi¾?loglogVB/E I/Os by Arge et al. for all values of V, E and B, for which I/O optimality is still open. In particular, our algorithm matches the lowerbound \varOmega E/V \cdot \mathrm{Sort}V$$ΩE/Vi¾?SortV, when E/V \ge B^{\epsilon }$$E/Vi¾?B∈ for a constant \epsilon > 0$$∈>0, an O\log \log B$$OloglogB factor improvement over the algorithm of Arge et al. Our algorithm can compute the connected components too, for the same number of I/Os, which is an improvement on the best known upper bound.
Evacuation Planning of Large Buildings Using Ladders
Lecture Notes in Computer Science, 2012
Evacuation planning of a building in case of an emergency has been widely discussed in literature... more Evacuation planning of a building in case of an emergency has been widely discussed in literature. Most of the existing approaches consider a building as a static graph with fixed, predefined exits. However, in severe disaster situations, it is desirable to create additional exits for evacuation purposes. A simple and practical way of creating additional exits is to place ladders at those locations that can reduce evacuation time effectively. For large buildings, finding optimal locations for a limited number of available ladders to utilize them effectively is not possible without using any systematic approach.
I/O efficient algorithms for the minimum cut problem on unweighted undirected graphs
Theoretical Computer Science, 2015
ABSTRACT
External Memory Soft Heap, and Hard Heap, a Meldable Priority Queue
Lecture Notes in Computer Science, 2012
ABSTRACT
Modeling of building evacuation using ladders
Fire Safety Journal, 2013
Existing research and tools used for building evacuation planning do not take into account the co... more Existing research and tools used for building evacuation planning do not take into account the common practice of using ladders for rescue operations. The ladders provide a simple and practical way of creating additional exits with a potential to significantly reduce the evacuation time. Use of ladders is critical in case of severe disasters and when some normal exits get blocked. To our knowledge, this paper presents the first systematic planning approach for optimal placement of a limited number of available ladders. We first propose modifications to the existing models of buildings to incorporate ladders and ladder points (locations in the building where ladders can be placed). Next, we develop optimization formulations to solve the following evacuation planning problems when a limited number of ladders are available: (i) optimal deployment of a given number of ladders to minimize evacuation time and (ii) finding minimum number of ladders along with their locations necessary to evacuate a building in a given evacuation time. These problems, posed as integer linear programming formulations, are applied to two case studies to demonstrate the importance of using ladders. The results show that evacuation plans can be significantly improved by optimally placing ladders at the selected ladder points.
Fire Safety Journal, 2013
The publisher regrets that the printed version of the above article has two corrections and they ... more The publisher regrets that the printed version of the above article has two corrections and they are listed below: (1) In Fig. , at node N4, label should be (0, ∞) instead of (0,b). (2) The correct version of Fig. is given below: The publisher would like to apologise for any inconvenience caused.
Indoor Evacuation Planning
Geospatial Infrastructure, Applications and Technologies: India Case Studies, 2018
Making efficient evacuation plans for large buildings is an important task in operating a large b... more Making efficient evacuation plans for large buildings is an important task in operating a large building infrastructure. The plans require a capacitated building network, well-defined exit points, and expected occupancy of a building to analyze evacuation time in different scenarios. Such type of analyses helps us in both the re-design to improve evacuation times as well as for creating more optimal allocations of the building spaces. In this chapter, we model the problems in the following two different types of situations and present heuristic algorithms which are extensions of the existing capacity-constrained route planner (CCRP) algorithm: (i) Additional exits with ladders need to be placed for faster evacuation, and (ii) all evacuees cannot be evacuated by themselves and require help from rescuers. For large and complex buildings, it is difficult to model indoor network and understand an evacuation plan for different scenarios without any user interactive interface. We have designed an evacuation planner which is publicly available and can be used for making indoor network from floor images, creating multiple scenarios, computing evacuation plans of these scenarios, and performing sensitivity analyses of an evacuation plan to evacuees’ behavior.
Mining Swarm Patterns in Sliding Windows over Moving Object Data Streams
Proceedings of the 25th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems, 2017
Several emerging applications such as traffic management and urban emergency response systems oft... more Several emerging applications such as traffic management and urban emergency response systems often need to identify groups from recent window of the moving object data. These requirements mandate algorithmic solutions that are time and memory efficient for adding new data incrementally and removing stale data. In this paper, we consider the problem of finding closed Swarms over a sliding window. The key challenges in computing closed Swarms over a sliding window are: (i) Search space for computing closed Swarms from new data is large (ii) Removal of old data leaves many non-closed Swarms which need to be identified and deleted. None of the existing methods are efficient in adding new data and removing old data for large datasets. This paper presents an efficient incremental graph based method for computing Swarms over sliding windows. We use a real dataset to show the performance of our method. The complexity analysis as well as experimental results demonstrate that our method is significantly faster than the existing incremental method over sliding windows with increased memory requirement. In particular, our method is shown to be 7-13 times faster with 3-5 times memory overhead in all the experiments.
A Distributed Learning Simulation Platform for Edge Hierarchies
2020 International Conference on COMmunication Systems & NETworkS (COMSNETS), 2020
We develop a distributed learning simulation platform that allows users to create multi-level Edg... more We develop a distributed learning simulation platform that allows users to create multi-level Edge hierarchy for a given application by simulating resource constrained Edge devices and communication links amongst them. The resulting Edge computing hierarchy is used to run a given DNN for the application in a data distributed fashion, rolling up learned parameter values up a hierarchy of parameter servers that merge parameters received from the lower levels. The root of this hierarchy has the latest model, which is then pushed lazily back down the tree to the Edge servers. The platform can be used to study cost vs accuracy analysis of a given application for different Edge hierarchy configurations. We use handwritten digit recognition problem as a case study to show the usefulness of our platform.
Mining Swarms from Moving Object Data Streams
Current methods for mining groups from moving object data work with entire data stream. However, ... more Current methods for mining groups from moving object data work with entire data stream. However, there are several emerging applications such as traffic management and urban emergency response systems which often need to identify groups from recent window of the moving object data. These requirements mandate algorithmic solutions that are time and memory efficient for adding new data incrementally and removing stale data. In this paper, we consider the problem of finding closed swarms over a sliding window. Large search space for computing closed swarms from the new data is the main key challenge in computing closed swarms over a sliding window. None of the existing methods are efficient for this. This paper presents an efficient incremental graph-based method for computing swarms over sliding windows. We demonstrate the performance of our method on two real datasets. The results show that our method is significantly faster than the existing incremental method over sliding windows w...
Proceedings of the 1st ACM SIGSPATIAL International Workshop on the Use of GIS in Emergency Management - EM-GIS '15, 2015
This paper focuses on indoor evacuation path planning problem where the objective is to find evac... more This paper focuses on indoor evacuation path planning problem where the objective is to find evacuation paths for each evacuee such that the evacuation egress time is minimized. Since paths are dependent on the distribution of evacuees, initial positions of evacuees are required to find optimal paths during emergency. Instrumenting a building to obtain initial positions and count of people in a building is very challenging (and costly), and hence evacuation plans are prepared for a few expected distributions. Generally, a standard plan based on a predominant distribution of evacuees is posted on the walls inside an indoor facility for people to follow. However, the actual distribution may be another distribution from among the possible distributions. In this work, we consider the problem of finding the distribution that prevails at the evacuation time so that evacuees can be guided to follow the optimal paths (leading to minimum time) rather than following the standard plan. We propose a cost-effective solution to this problem by observing movement of people within a specified time period, labeled distribution detection window, using minimum number of optimally located sensors. This is in contrast to existing approaches which assume that unlimited sensors are available to instantaneously obtain the exact distribution of evacuees at the time of incidence. To our knowledge this paper presents the first formal evacuation planning approach that enables the user to optimally tradeoff the delay in distribution detection with the cost of the deployed sensor network used to obtain this distribution. Our approach is based on the popular heuristic denoted as Capacity Constrained Routing Planner (CCRP). Our approach is illustrated by a set of experiments on two case studies. The results demonstrate that evacuation plans obtained using minimum number of sensors are better than the standard plans and are comparable to evacuation plans computed using unlimited number of sensors.
An I/O Efficient Algorithm for Minimum Spanning Trees
Lecture Notes in Computer Science, 2015
An O\mathrm{Sort}E\cdot \log \log _{E/V} B$$OSortEi¾?loglogE/VB I/Os algorithm for computing a ... more An O\mathrm{Sort}E\cdot \log \log _{E/V} B$$OSortEi¾?loglogE/VB I/Os algorithm for computing a minimum spanning tree of a graph G=V,E$$G=V,E is presented, where \mathrm{Sort}E=E/B\log _{M/B}E/B$$SortE=E/BlogM/BE/B, M is the main memory size, and B is the block size. This improves on the previous bound of O\mathrm{Sort}E \cdot \log \log VB/E$$OSortEi¾?loglogVB/E I/Os by Arge et al. for all values of V, E and B, for which I/O optimality is still open. In particular, our algorithm matches the lowerbound \varOmega E/V \cdot \mathrm{Sort}V$$ΩE/Vi¾?SortV, when E/V \ge B^{\epsilon }$$E/Vi¾?B∈ for a constant \epsilon > 0$$∈>0, an O\log \log B$$OloglogB factor improvement over the algorithm of Arge et al. Our algorithm can compute the connected components too, for the same number of I/Os, which is an improvement on the best known upper bound.
Evacuation Planning of Large Buildings Using Ladders
Lecture Notes in Computer Science, 2012
Evacuation planning of a building in case of an emergency has been widely discussed in literature... more Evacuation planning of a building in case of an emergency has been widely discussed in literature. Most of the existing approaches consider a building as a static graph with fixed, predefined exits. However, in severe disaster situations, it is desirable to create additional exits for evacuation purposes. A simple and practical way of creating additional exits is to place ladders at those locations that can reduce evacuation time effectively. For large buildings, finding optimal locations for a limited number of available ladders to utilize them effectively is not possible without using any systematic approach.
I/O efficient algorithms for the minimum cut problem on unweighted undirected graphs
Theoretical Computer Science, 2015
ABSTRACT
External Memory Soft Heap, and Hard Heap, a Meldable Priority Queue
Lecture Notes in Computer Science, 2012
ABSTRACT
Modeling of building evacuation using ladders
Fire Safety Journal, 2013
Existing research and tools used for building evacuation planning do not take into account the co... more Existing research and tools used for building evacuation planning do not take into account the common practice of using ladders for rescue operations. The ladders provide a simple and practical way of creating additional exits with a potential to significantly reduce the evacuation time. Use of ladders is critical in case of severe disasters and when some normal exits get blocked. To our knowledge, this paper presents the first systematic planning approach for optimal placement of a limited number of available ladders. We first propose modifications to the existing models of buildings to incorporate ladders and ladder points (locations in the building where ladders can be placed). Next, we develop optimization formulations to solve the following evacuation planning problems when a limited number of ladders are available: (i) optimal deployment of a given number of ladders to minimize evacuation time and (ii) finding minimum number of ladders along with their locations necessary to evacuate a building in a given evacuation time. These problems, posed as integer linear programming formulations, are applied to two case studies to demonstrate the importance of using ladders. The results show that evacuation plans can be significantly improved by optimally placing ladders at the selected ladder points.
Fire Safety Journal, 2013
The publisher regrets that the printed version of the above article has two corrections and they ... more The publisher regrets that the printed version of the above article has two corrections and they are listed below: (1) In Fig. , at node N4, label should be (0, ∞) instead of (0,b). (2) The correct version of Fig. is given below: The publisher would like to apologise for any inconvenience caused.