Stage 2 presentation
Transcription
Stage 2 presentation
Street Traffic Estimation System Ashish Dhar Motivation Traffic jams & congestion prevalent in most cities What will it do? A traffic monitoring system required Estimate the state of traffic on roads What will be its benefits? Will enable quick reaction measures Change time of traffic light signals Commuters take alternative routes Planning a better road network The Key Difference Unstructured Traffic Our Focus Structured Traffic Previous Work Focus The Indian Traffic Scenario Traffic is unstructured, lack of lane discipline Chaotic & dense traffic, Lot of maneuvering Difficult for video, radar, infrared & laser based systems These assume some structure in traffic Magnetic loops & sensors require traffic in lanes Impossible to count vehicles & estimate their speeds Numerous type of vehicles Different types vehicles move together on the road Thinly spaced Difficult to detect & classify vehicles Problem Statement To estimate traffic patterns in India scenarios Low cost, easily deployable Specifically, Volume of traffic (traffic density) Percentage of road occupied by vehicles Movement pattern Stationary, stop & go, slow or fast moving System Overview Sensing Unit Maximum ADC count without any external magnetic field ! Shifting the Mid-point Voltage Resolving out of range voltage issue Stepped down mid point voltage to 1.25V Inserted a voltage divider Need to Improve sensitivity & detection range Set/Reset Circuit Integrated with Micro Controller Testing & Verification of Hardware Objective: To verify the working of our sensing unit Test signals for different vehicles at different distances, speed Vehicle Signature Sampling rate = 20Hz, Distance ~ 1.75m, vehicle = passenger car It is possible to get vehicle signatures at standoff distance Additivity Test Objective: To verify additive nature of magnetic signal Distance of sensor from: Can we employ amplitude as a measure of traffic density? Minibus ~ 3.5 meter Autorickshaw ~ 1 foot Sampling rate = 20 Hz Additivity Test Magnetic signal of a moving vehicle is additive to static one Stationary vehicle results in static magnetic sensor value Does stationary traffic also give static readings? Experiment to Study Correlation Between Traffic Pattern & Sensor Readings Sampling rate = 40Hz, height = 1m, distance = 1 foot Readings timestamped to one second resolution Video camera used to record ground truth (state of traffic) Methodology Two experiments of duration 14 & 11 mins All samples in 1 sec time window averaged Magnetic sensor readings logged at base station Removed quantization error & smoothened plots Ground Truth Determine traffic patterns by watching videos Manually annotated traffic patterns Group plots of similar traffic pattern for analysis Magnitude= Magnetic Sensor Signal Magnitude= ADC0 2 ADC12 Static Traffic Characteristics Stationary traffic results in constant magnitude value at a deviation from base reading Different level of deviations with varying type and position of vehicle Cannot use signal amplitude as a measure of traffic density for static traffic Partially Stationary Traffic Small vehicle close to the sensor gives significant deflection Can result in false negative for stationary traffic if vehicles (2, 3 wheelers) keep crawling ahead Far Lane Effect Large vehicles on far side of the road give same magnetic signal as small vehicles that are close to the sensor Moving Traffic Characteristics Inter vehicle spacing & speed impact moving traffic magnetic signal Observations Can identify no traffic condition Can identify stationary traffic Stable magnitude values at base readings Stable value at some deviation from base readings Small vehicles close to the sensor distort stable waveform Maximum amplitude not an indicator of traffic density for static traffic Moving traffic patterns difficult to capture with 1 magnetic sensor Distinct peaks can be seen as a measure of inter vehicles spacing Maximum amplitude as a measure of traffic density for moving traffic Width, or rate of change of prominent waveforms as a measure of traffic speed Conclusion Built & tested traffic sensing unit Interfaced wireless sensor node with magnetic sensor Conducted experiments to establish correlation between sensed data & actual traffic Determined relationship between static traffic & plot characteristics Identified problem of near & far lane effects Identified some characteristics of sensor signal that can be used to estimate moving traffic patterns Future Work Benchmarking experiments Vehicle speed v/s signal amplitude & width Inter vehicle spacing v/s signal waveform Additivity Tests (both vehicles moving) Vehicle distance v/s signal amplitude Hardware upgradation Conduct experiments for more varied traffic conditions Build & test traffic estimation module Extra Slides “Composite” Signal Effect Our magnetic sensor is at standoff distance Picks up signals from multiple vehicles at different distances and moving with different speeds Received signal is “composite” of several signals Deformed magnetic signal waveform Need to map “composite” magnetic signal to specific traffic patterns Utility SMS based services that alert users about congestion Automatic traffic light timers GIS systems that suggest less congested paths Analysis tools that help manage traffic and plan extensions to the road network Why Current traffic Technologies Do Not Solve Our Problem Technology Traffic parameters observed Infrastructure requirements Drawbacks Magnetic Loops Count vehicles moving on lanes Requires digging up of roads Count vehicles moving on lanes Expensive intrusive Camera Based Systems Microwave Radar Laser based systems Infrared based system Detect, count & classify vehicles Can be mounted on poles along Low performance in poor visibility Do not require laned traffic roads Needs to be secured as chances of theft higher Large vehicles obscure smaller ones Cannot detect stationary vehicles Needs overhead structures Overestimating speed & occupancy Estimate speed Digging up the road not needed values specialized expensive equipment Laned traffic Count classify estimate speed of vehicles Needs overhead structure Assumes structured traffic Digging up the road not needed specialized expensive equipment Can only detect vehicles Expensive (cheaper than loops) Fog, Snow and rain degrade performance Assumes structured traffic Ultrasonic Detectors Anisotropic Magneto Resistive Sensors Detection only. Speed detection ones are expensive Overhead mounting points needed Detect Classify and estimate speed of vehicles Little digging Assumes structured traffic Expensive Cannot work in dense, varied traffic AMR Magnetic Sensor Able to detect disturbance in earths magnetic field on their sensitive axis Stationary or moving vehicle create this disturbance Commercial AMR Sensors have 1, 2 or 3 axis Advantages Depending on its ferrous content Can manufactured in bulk at low cost mounted in commercial IC packages Small in size, highly sensitive, immune to noise and reliable Exponential drop in detection ability with distance AMR Magnetic Sensor for Traffic Standoff Flux Density Monitoring Distance (foot) (milligauss) Detecting Vehicles Estimating Speed Use simple threshold based mechanism Use two sensors in quick succession at a fixed distance Time difference between peaks used to calculate speed 1 270 3 75 5 10 10 2 12 <1 A typical automotive magnitude (flux density) versus sensor standoff distance Classification ? Ferrous object disturbance in uniform field Source: Vehicle detection and compass applications using amr magnetic sensors How is AMR Sensor Used for Vehicle Classification A vehicle can be considered to be a composite of dipoles Forward Depending upon where the ferrous content is present Vehicle Each vehicle has a unique signal on the magnetic sensor This is exploited for vehicle classification Magneti c SensorX Y Z 1 ft. above the ground A Typical magnetic Sensor deployment & its orientation Vehicle Classification Using AMR Sensors Earth’s field variations for a van driving over an AMR magnetic sensor Earth’s field variations for a car driving over an AMR magnetic sensor Source: Vehicle detection and compass applications using amr magnetic sensors Traffic Density Percentage of a road area occupied by vehicles Higher traffic density implies congestion Map sensor deflection to approximate traffic density Deflection proportional to vehicle’s iron content Iron content proportional to vehicle’s size