Modeling network of vehicles in a heterogeneous non-urban environment


Last week's Friday Seminar featured EECS Professor Ruzena Bajcsi and grad student Katie Driggs-Campbell presenting their research on developing driver models to address issues in heterogeneous environments (with autonomous vechiles and human drivers). Their research group has focused on the relationship of the individual driver in the environment around them. Modeling driver behavior is important because driver distraction makes drivers unpredictable and leads to significant safety implications. NHTSA reports in 2013, 424,00 people were involved in distracted driving crashes with 3,154 fatalities. Even with the rapid advances in automated and connected vehicle technology,  Bajcsy forseees heterogeneous environments for at least the next 15 years, so the ability of the models controlled the autonomous vehicles will need to understand potential behaviours of human drivers. To develop the models using control theory, Driggs explained how they used human-in-the-loop driving simulations to better detect different stages of driver distraction and maneuvering. These models can then be used for semiautonomous vehicular control

Stay tuned for this Friday's Seminar, which is the last of the semester. ITS gradstudent Haoyu Chen will present his dissertation research on improving public transit at city-wide scales. 

Monitoring Traffic for Incidents and Extreme Congestion Events

flickr photo shared by canihazit under a Creative Commons ( BY-NC-ND ) license

Last week's Friday Seminar welcomed back ITS alumn, now Univeristy of Illinois Urbana-Champaign Assistant Professor, Dan Work. He presented some of his work in the area of monitoring traffic data to better responde to incidents and extreme congestion events. The talk began with focus on how his group uses the app TrafficTurk for low-cost traffic sensing, which has been used in managing big events like homecoming weekend at University of Illinois and the Farm Progress Shows in Decatur, IL (where over 100,000 people flock for the event). He then talked about how he's used this sensing data to more quickly respond to traffic incidents through multiple model particle filtering. He concluded the talk preseting some preliminary results from New York City taxi data (which is shared here as open data) on the congestion in the city following Hurricane Sandy. The data shows that there was not much congestion during evacuation, as there was a system and a plan in place, but that the extreme congestion after the storm shows a need for better post-event planning and coordination. 

You can find more of Work's publications (and source code) here

Oregon's New Opt-In Mileage Tax Pilot

Interstate 5 from the Portland Aerial Tram

This week Oregon DOT (ODOT) announced the public trial of a proposed new mileage tax. They're calling it a Road Usage Charge Program and it will beging July1, 2015. They are looking for 5,000 volunteer drivers to launch the program, which will be trailblazing user-based fees like this. 

New funding structures for transportation are needed as it's not clear how much longer the federal gas tax will last, which is especially problematic given the depletion of the Highway Trust Fund. To cover the challenging gap between current funding and what is needed to maintaining the current U.S. transportation system, new finding models are emerging. Road pricing, often implemented as tolling, is a very common method. Vehicle Miles Traveled (VMT) based taxes are another user-based method which is being discussed nationally, though Oregon is the first state to take steps to implement such a tax. Critics of these methods argue that such taxes penalize fuel efficient vehicles, or that they are regressive taxes and social inequities must be accounted for. The equity question is currently a very active research topic.  

This past spring, a mileage tax for California drivers was introduced in the state Senate. SB-1077 was voted by the Senate and Assembly, and approved Governor Jerry Brown in September. 

Institute of Transportation Studies Friday Seminar: Lane Changing - Mysteries on Behavior and Modelling

Changing lanes

This week's Friday Transportation Seminar is about ellusive lane changing behaviors. Victor Knoop, Assistant Professoor of Transport & Planning at TU Delft will present, "Lane Changing - Mysteries on Behavior and Modelling."

Traffic congestion often is related to lane changes - at a lane drop bottleneck, on ramp, or weaving section. It is therefore essential to have a good description of the lane change maneuvers performed by drivers. Whereas much attention has been given to car-following behavior (how much distance do people keep), lane changing did not get the same amount of attention. This seminar will touch upon three aspects related to lane changing. Firstly, a large-scale data analysis shows that simple concepts of trying to go to a faster lane combined with gap accepting does not provide a satisfactory model. In the seminar outcomes of the data analysis are shown at it is discussed what can be learned from it. Secondly, calibration and validation are required for any model, but there is no standardized method for calibrating lane-change models. It will be shown that the methods have been chosen carefully. Thirdly, the differences in driving strategy between drivers with regard to lane changing are discussed.

The Friday Transportation Seminar takes place on September 19, 2014 from 4:00-5:00 PM in 290 Hearst Memorial Mining Building. Cookie Hour immediately precedes it at 3:30 PM in the same location. (Note: Cookie Hour is not in the library!) There will also be a no-host Happy Hour at LaVal's at 5:00 PM.

In California, self-driving cars need a permit.


Yesterday California's 3-foot passing law for cars and bicycles took effect. New regulations from the California DMV about Autonomous Vehicles in California also took effect - all self-driving cars will need a permit for testing. Audi, Mercedes, and Google have already applied for an received the first set of permits in the state. (If you're keeping tabs, Audi received the first permit.)

Regulations and licenses for operation of autonomous vehicles in California will be finalized by January 1, 2015

For background on policy and regulation of autonomous vehicles, check out the 2014 RAND report "Autonomous Vehicle Technology: A Guide for Policymakers," which builds upon concepts from the 2009 PATH report "Liability and Regulation of Autonomous Vehicle Technologies."

Institute of Transportation Studies Friday Seminar: A sensor-based and spatially-enabled system for next generation Intelligent


Friday's Transportation Seminar is all about smart infrastructure. James Tsai from Georgia Tech presents, "A sensor-based and spatially-enabled system for next generation Intelligent and sustainable infrastructure management."

Roadway infrastructures, including pavements, bridges, and signs are deteriorating rapidly due to material aging, improper usage, harsh environments, and damages resulting from natural or man-made hazards. With the advancement of sensor technologies, it become feasible to collect the large-scale in-field detailed infrastructure data, such as 3D pavement surface data, using high-performance cameras, lasers, LiDARs, and Inertial Navigation System (INS) to gain better insight understanding of the large-scale in-filed infrastructure behavior. This talk first presents a framework for the sensor-based and spatially-enabled next generation Intelligent and sustainable infrastructure management system, including the key components of data acquisition, automatic information extraction, data integration, and intelligent infrastructure management. An intelligent sensing system has been developed, using 2D Imaging, Laser, LiDAR, and GPS/GIS Technologies with artificial intelligent and pattern recognition to automatically detect pavement surface distress, including rutting, cracking, raveling, etc. along with their detailed level characteristics for determining pavement health condition. The availability of high-resolution roadway images, 3D pavement surface data, and 3D LiDAR data has brought us a great opportunity and new challenges. This calls for a new concept to model this detailed level of big data for revealing new values for infrastructure management. First, we need to effectively extract valuable decision-support from this big data. For cracking, an innovative crack fundamental element (CFE) model that is a topological representation of cracks to support crack classification, diagnosis, and intelligent pavement management will be presented; this CFE provides researchers a mathematical foundation for modeling large-scale, in-field pavement/infrastructure crack characteristics to study crack propagation behavior at multiple scales will be presented. Examples of developing an innovative and sustainable pavement preservation method and developing intelligent crack sealing planning using emerging sensor technologies will also be presented.

The Friday Transportation Seminar takes place on September 12, 2014 from 4:00-5:00 PM in 290 Hearst Memorial Mining Building. Cookie Hour immediately precedes it at 3:30 PM in the same location. (Note: Cookie Hour is not in the library!) There will be a no-host Happy Hour at LaVal's at 5:00 PM.

Institute of Transportation Studies Friday Seminar: Large-Network Travel Time Distribution Estimation for Ambulances

Ambulance in Georgetown. BW.

It's almost Friday, so it's almost time for the Friday Transportation Seminar. Remember that this semester the seminars and Cookie Hour are in 290 Hearst Memorial Mining Buidling. We also encourage you to follow along (and participate) with the Twitter hashtag #itsberksem

This week's Friday Seminar features Dawn Woodward from Cornell presenting, "Large-Network Travel Time Distribution Estimation for Ambulances."

We present methods to predict the time required for an ambulance to drive to the scene of an emergency. This forecast is critical for deciding how many ambulances should be deployed at a given time, where they should be stationed, and which ambulance should be dispatched to an emergency. Specifically, we predict the distribution of lights-and-sirens ambulance driving time on an arbitrary route in a road network, using automatic vehicle location data and trip information from previous ambulance trips. We train a statistical model using a computationally efficient procedure; challenges include the large size of the network and the lack of trips in the data that follow the route of interest. We demonstrate the operational impact of our methods using data from Toronto Emergency Medical Services, and discuss ongoing efforts to incorporate our methods into a software package used by ambulance services.

The Friday Seminar takes place on September 5, 2014 from 4:00-5:00 PM in 290 Hearst Memorial Mining BuildingCookie Hour immediately precedes it at 3:30 PM in the same location. (Note: Cookie Hour is not in the library!) There will be a no-host Happy Hour at LaVal's at 5:00 PM.

Institute of Transportation Studies Friday Seminar: Lessons Learned from Spatiotemporal Studies of Freeway Carpool Lanes


Late August means the end of summer is nigh, students are back, classes are in session, and it's the return of the Friday Transportation Seminars. This semester there are some changes to the seminars - Cookie Hour and the seminar will take place in the same location in 290 Hearst Memorial Mining Buidling. We also encourage you to follow along (and participate) with the Twitter hashtag #itsberksem

This week's Friday Seminar features ITS's own Professor Michael Cassidy presenting, "Lessons Learned from Spatiotemporal Studies of Freeway Carpool Lanes."

The presentation explores how the segregation of distinct vehicle classes on a roadway can improve travel conditions for all of the classes. Insights come using freeway carpool lanes as case studies. Spatiotemporal study of real sites shows (i) how the activation of a continuous-access carpool lane triggers reductions in vehicle lane-changing maneuvers, and (ii) how the reduced lane-changing can “smooth” and increase bottleneck discharge flows in a freeway’s regular lanes. Theoretical analysis predicts that, thanks to this smoothing effect, even underused carpool lanes can diminish both the people-hours and the vehicle-hours traveled on a freeway. Relevance to bus lanes is briefly discussed. Further insights come via critiques of certain practices that degrade the effectiveness of carpool lanes. Spatiotemporal traffic data reveal that a policy aimed at improving carpool-lane speeds has backfired, owing to a friction effect. The policy mandates the eviction of select fuel-efficient hybrid vehicles from carpool lanes. These evictions have caused queues to expand in regular lanes during the rush. And these expanded queues, in turn, slow vehicles in the adjacent carpool lanes. Spatiotemporal data further show that efforts to combat the friction effect by deploying limited-access carpool lanes can also backfire, because the designs for these lanes are prone to creating bottlenecks.

The Friday Seminar takes place on August 29, 2014 from 4:00-5:00 PM in 290 Hearst Memorial Mining Building. Cookie Hour immediately precedes it at 3:30 PM in the same location. (Note: Cookie Hour is not in the library!) There will be a no-host Happy Hour at LaVal's at 5:00 PM.

Celebrating Independence Day!

Where baby freeways come from

Tomorrow is Independence Day here in the United States, whi means the library (and the whole UC Berkeley campus) will be closed in observance of the Federal holiday. 

It also means holiday traffic as people head out of town for a long weekend, outdoor recreation, or to fireworks displays.  Holiday weekends tend to see an over-representation of fatal and injury crashes on the roadways, largely due increased drunk driving. Increased impaired-driving enforcement has helped reduce crash fatalities in the United States. Checkpoints could be even more effective with location optimization modeling

So have a festive and safe weekend. Don't drive impared. We'll see you next week!

New NCHRP Synthesis: Response to Extreme Weather Impacts on Transportation Systems

345/365 Flood

This week a new NCHRP Synthesis was published by TRB that covers the effects of extreme weather incidents, such as Hurricane Sandy, on transportation systems. NCHRP Synthesis Report 454: Response to Extreme Weather Impacts on Transportation Systems provides background on the issue and the current state of the practice. The full report can be read here

Syndicate content