Potential Risks and Mitigation Strategies

At this time we feel more confident in our solution and we were able to finalize most of our solution approach with specific hardware and software we are using. Earlier in the week we were hesitant on how we can create training data and test our optimization model effectively, however we found a software that allows us to simulate scenarios and retrieve the data in Python. This software is called SUMO (see Kaitlyn’s Report for more details).

Changes to System Design

Our schedule has changed in regards to the optimization algorithm, since we learned about using SUMO and are now using it for our simulation. This resulted in us switching the simulation before the optimization algorithm is developed, however we are still on track for our timeline overall.

No other changes to the schedule have been made.

Overall Takeaways and Progress

• Github repo setup
• Finalized optimization algorithm – deep q-learning
• Finalized image detection algorithm – Haar cascade
• Received ReoLink IP camera to start testing
• Researched RPi WiFi networking
• Worked on Design Review

Needs Our Solution Meets

Public Health, Safety, Welfare (Ankita)

Part A: … with respect to considerations of public health, safety or welfare. Note: The term ‘health’ refers to a state of well-being of people in both a physiological and psychological sense. ‘Safety’ is the absence of hazards and/or physical harm to persons. The term ‘welfare’ relates to the provision of the basic needs of people.

Traffix will address the concern of public health, safety, and welfare by potentially reducing the number of collisions at an intersection. By optimizing traffic intervals, we can lower the risk of impatient drivers running red lights and pedestrians crossing when it is not safe for them to do so. We can also improve overall public health by making traffic less of a hassle for the typical Pittsburgh commuter.

One thing to note, though, is that our system must provide intervals that are long enough for cars and pedestrians to cross the intersection safely — in other words, we cannot “overoptimize” the system. In that sense, our product may also introduce safety concerns, which is why it must undergo a rigorous testing stage with multiple simulated scenarios.

Social Factors (Zina)

Part B: … with consideration of social factors. Social factors relate to extended social groups having distinctive cultural, social, political, and/or economic organizations. They have importance to how people relate to each other and organize around social interests.

According to data from the USDoT’s Bureau of Transportation Statistics, Americans take over 100 billion trips in cars every month. Given this fact, it’s clear that personal vehicle ownership and use has become an integral part of the daily routine of your typical American citizen. In a culture so dependent on driving, even minor inefficiencies in the current traffic control systems can have a significant impact on a variety of social factors. 

Sub-optimal light timing intervals at intersections leave people waiting for excessive amounts of time. While idling, most cars are not designed to temporarily shut the engine off, which cumulatively can waste immense amounts of fuel. Not only is this bad for people’s wallets, but the unnecessary emissions are also harmful for the environment in which we all live. Besides wasting fuel, waiting for inefficient lights to change wastes another valuable resource: people’s time. Even if it’s only a few minutes over the course of a day, this can seriously start to add up when driving is something you have to do every day to commute to work, take your kids to school, or go buy groceries. Traffix aims to greatly reduce the severity of these inefficiencies that result in wasted time and money.

Economic Factors (Kaitlyn)

Part C: … with consideration of economic factors. Economic factors are those relating to the system of production, distribution, and consumption of goods and services.

We believe that our solution will economically benefit drivers due to a decrease in traffic accidents. On average, accidents cost the US $340 billion a year, so minimizing accidents will lead to less medical costs related to injuries. Studies show that increased traffic congestion leads to increased likelihood of car crashes, so our solution will ultimately save users money in terms of minimized traffic accidents.

Additionally, our solution saves city drivers money in terms of time spent in traffic. Data suggests that in 2018, Pittsburgh was the #7 most congested urban area and on average costs $1,776 per driver due to time lost in congestion and $1.2 billion for the city. By optimizing traffic flow by even 10%, which was our target metric, we can reduce costs for the city by millions with widespread implementation.

Potential Risks and Mitigation Strategies

The main risk we currently foresee is being unable to get the IP cameras set up at an actual intersection to send data to the Raspberry Pi. We have a plan in mind for this (detailed in Ankita’s Status Report), but if it doesn’t work we have a couple of contingency plans:

• Run our OpenCV object detection script on one side of the intersection with a wired camera connection, and simulate the other 3 sides for the optimization algorithm.
• Scrap the cameras altogether and just use past traffic camera footage for demo purposes; demo the hardware and software of our system separately and not as one cohesive system.
• Adjust the scope of our project to a miniaturized intersection setup with toy cars and pedestrians. If we end up doing this, we will need to change our schedule to accommodate for the construction of the model. We will also need to change the object detection algorithm significantly.

Changes to System Design

We don’t have any changes to the design as of yet, but we expect to make some in the next week as we finalize our camera setup strategy. If we have to switch to one camera and simulate the other three sides of the intersection, we would be sacrificing our testing metrics; instead of testing our system against a real world situation, we would have to simulate what the real world light timings would be (assuming a fixed-time interval). This would not be as robust a comparison. Scrapping the cameras altogether would mean we would not be able to demonstrate a fully integrated system, which would also take away from our use case requirements. Finally, using a miniaturized intersection setup will require us to replan a lot of our schedule, as creating the model will take up a lot of time.

As of yet, our schedule is unchanged.

Overall Takeaways

The project seems to be on track according to our planned schedule but we will need to ramp up work in the next couple of weeks to get the software working while the camera system gets figured out. Integration will take a great deal of time, so keeping as much slack time as possible is optimal.