While self-driving cars have established their presence in the automotive industry, we are far from achieving full automation. Drivers nowadays can control acceleration, braking, and steering partially, meaning they are still required to be engaged while driving. Advancements in technology have greatly impacted and influenced vehicles of this day and age whether positively or negatively, it can be concluded that we haven’t fully trusted this technology to act completely for us in our everyday driving scenarios. Although our technologies are capable of producing amazing results for what knowledge, construction, and research we have as of now, the technologies we have, aren’t actually up to speed in their capabilities. It is evident that self-driving cars driving on the road today aren’t completely self-driving, resulting in the driver still acting as a catalyst to the vehicle’s self-driving mode, and if anything were to occur, the self-driving mode would be turned off, but the biggest issue when it comes to self-driving cars is it’s safely and how we would be able to approach ethical issues that would relate to self-driving cars. This topic has been discussed in source 1 by Jason Borenstein, Joseph R. Herkert, and Keith W. Miller in their article “Self-Driving Cars and Engineering Ethics: The Need for a System Level Analysis”, in which they’ve discussed multiple and major concerns there would be if companies rush to bring self-driving vehicles into the market because it many cases it would result in compromising consumer safety and autonomy. As a result of this concern, they have described and discussed two traffic scenarios with the acknowledgment that these aren’t the only possible scenarios, but that they are highlighting and showing some of the key system issues and potential resolutions. Relating to the concern of self-driving cars failing to deal with certain emergencies of the vehicle’s inability to detect the road, which can lead to major accidents involving other vehicles or pedestrians. Thus, source 2 of the article “A Co-Point Mapping-Based Approach to Drivable Area Detection for Self-Driving Cars” by Ziyi Liu, Siyu Yu, and Nanning Zheng, proposes a self-adaptive method for drivable area detection to acquire a safe and structured future for self-driving cars to be roaming through our streets. 

Source 1 dives into the subject of technology and its impact on human lives and society by discussing a “system level of analysis that engages with the interactions and effect that these cars will have on one another and on the socio-technical systems in which they are embedded.” By making use of two traffic scenarios, highlighting some of the problems that policymakers and designers should consider when relating to technology, they can describe three approaches that can address those complexities. Scenario 1 demonstrates two vehicles, one planning on exiting the highway and one planning on merging onto the highway, however, ended up being completely parallel to one another. With this, there could be difficulties in knowing if the two vehicles would be able to communicate with one another, meaning that the technology itself and the software used for the vehicles may not be sufficient in coordinating a safe and effective maneuver. As of scenario 2, the involvement of the previous 2 cars in this scenario, combined with pedestrians, a motorcyclist, a cyclist, and a random person driving a regular automobile, would greatly conclude that from all these different interactions amongst such entities would become a hazardous and messy scene to unfold. Source 2, follows the reasoning for source 1, which is fixing problems by finding a solution. In connection with source 1, the Engineers Proposed a method that was inspired by human driving behavior since computers don’t process situations with emotions. Both sources hold the same question and both acquire solutions for their problem. Each with evidence to support their methods and hypothesis, coming to a conclusion that can potentially fix or alter such problems revolving around self-driving cars.   

Taking to account the scenarios we would need to acquire a self-driving vehicle for the procedures that follow. It is advisable to contact an expert who will be in the vehicle, since self-driving cars still require the driver to pay full attention, if anything were to occur that would cause the self-driving mode to turn off, resulting in its default mode, where the driver will need to take control again. The procedure consists of having two vehicles perform a scenario, like in scenario 2, and analysis after the results. However, no matter the scenario, the self-driving cars in the experiment may not respond correctly to one another due, either “due to design differences and compounding variables such as the unpredictability of the driver in Car A.” Essentially, source 1 demonstrates a hypothesis, testing it, with multiple different factors that may influence the outcome and eventually, because technology follows commands and doesn’t possess judgment, any changes to the computer software in the vehicle would result in it slowing down, even halting, until the driver took over. Thus, this sudden shift from self-driving back to manual maneuvers can be dangerous and a hazard for other pedestrians and vehicles around you, as well as the driver. Source 2 also does a procedure, where they proposed a self-adaptive method, allowing for the vehicle to be able to adapt to its surroundings as a human would be “leveraging co-point mapping to fuse the pixel information from a monocular camera with the spatial information from a laser sensor.” The engineers from source 2 also located initial drivable areas by fusing obstacles classification, which results in image superpixels through self-learning. They used the Bayesian framework, a combination of information with an individual’s prior sense of probability to produce an outcome, to find the final drivable area. They tested their method using the ROAD-KITTI benchmark and at the end of the experiment, gathered their results, which ended up being that when comparing with others “fusing methods, demonstrated that the proposed method achieves state-of-the-art results without requiring training or assumptions about shape or height; this result validates our method as being robust and having a high generalization ability.” This procedure was carried out with materials that included 289 training images and 290 testing images. The experiments were evaluated from a bird’s eye view. They would document the results and place them in tables too, concluding the results.  

The results are as follows. Source 1, acknowledges that with various variables in play giving a direct and fixed answer would be quite impossible. In Source 1, “Self-Driving Cars and Engineering Ethics: The Need for a System Level Analysis” there wasn’t a section labeled as results, whereas in source 2, “A Co-Point Mapping-Based Approach to Drivable Area Detection for Self-Driving Cars” there was, labeled, experimental results and discussion. The reason for this is a wide majority of source 1 focuses on plausible scenarios that have an excruciating amount of variables that can happen at a given time, making the results unpredictable and incalculable. So, instead of giving factual, evident, and undisputable results, it instead answers multiple questions, gathering several different answers, as well as primarily giving a focus on the moral responsibility for computing artifacts or as they call it “The Rules.” The rules are enforcing a policy that if the products are manufactured, the engineers and the team that made those products would have a shared responsibility for the design, development, and deployment of self-driving cars. The enforcement of rules would be considered a practical way to encourage accountability. In the end, the result is arguments that push the reason for the ethical evaluation of self-driving cars to include system-level analysis of the interaction between vehicles. As for source 2, tables support their method and how it yields the best performance in PRE and FPR in the UMM and UU datasets. In the UMM database, their method with or without co-point and training set gathered the highest percentage with an average of around 91% beating out hydridCRF, mixedCRF, lidarHisto, and RES3D-Velo. Their method was able to cover the road area well. The tables were able to reveal the robustness of their method while being in different scenarios. Source 1 is quite bland in its rule sections, as it is just words, but source 2 conveys a more presentable and visual experience, giving the audience better indicators as to what they would be talking about. Although, I do understand and comprehend source 1 more easily due to its results not being contained from hypothesis and questions, source 2 usage of the tables and street images to help the reader visualize the drivable area for the self-driving car. 

From my gathering of the results between the two articles, I’ve examined both articles that even when different roads lead to the same goal of producing safe and responsive self-driving vehicles to be let on our roads. From the results gathered in source 2, the methods that they introduced beat the other popular methods, showing that by implementing their methods efficiency is backed up. Some things are plausible to do that can greatly affect the future of the automotive industry, especially the future of self-driving cars. The biggest difference between the discussion in each article was how source 1 interpreted questions that would be asked and answered, but still leaving plausible variables and interference, which doesn’t completely answer the topic. While in source 2 takes the data in their experiment and successfully proves how their method is superior. Though there are major differences there are similar traits between each source. Source 1 does have a figure and a table that helps to show the reader the experiment and procedure better. Source 2 contains 9 tables, while source 1 contains only 2. In source 2, “A Co-Point Mapping-Based Approach to Drivable Area Detection for Self-Driving Cars” table 1 contains data and percentages, but in source, “Self-Driving Cars and Engineering Ethics: The Need for a System Level Analysis”, it contains levels when it comes to the stages between self-driving cars with a person to a fully self-driven vehicle. Table 2 provides the U.S. federal checklist for autonomous vehicles. The data in source 2 is undisputed evidence, while the data in source 1 is primarily informational-based.  

Both articles end in their conclusion, reflecting upon their introduction while gathering and applying their results to showcase the benefits for the future of self-driving cars, in hopes it will not only impact human lives and society positively but remind us about the goals. However, source 1 concludes their introduction by linking back to the laws and regulations, rather than technologies, since they believe that many technical and ethical questions need to be answered before, “technologies such as V2V and Centralized Intersection Management become a reality”( Borenstein, R. Herkert, and W. Miller, 2017). Source 2 ends their conclusion by mentioning that it is necessary to implement their method to realize the real-time application of the method and its efficacy for self-driving cars. Source 1 leaves us questioning, while source 2 leaves us knowing. 

Although both reports resemble the same argument for improving and wanting to achieve a world where self-driving vehicles surface our roads safely, they apply different methods to reaching their audience and their goals. Even when each article takes a different approach to the future of self-driving cars, one focuses on the need for system-level analysis and rules, and the other focuses on driable area detection with their own better method, they both ultimately reach one goal.