Summary of “Warp Drive Research Key to Interstellar Travel”

In the article “Warp Drive Research Key to Interstellar Travel”, published on scientific american blog on 23rd April 2014, Mark Alpert writes about the current challanges in the research of interstellar travel.

He begins with the story of Zefram Cochrane, a fictional physicist of the Star Trek universe who invented the warp-drive in the year 2063 allowing the interstellar voyages of the starship Enterprise. This story leads to a real phyiscist, working in NASA’s Johnson space center in Houston, who is researching on the exact same topic. Harold “Sonny” White designed a tabletop experiment to create tiny distortions in spacetime. If his experiment succedes it could lay the foundation of a system that allows spacecrafts to sidestep the physical speedlimit of lightspeed. Instead of increasing the speed.of the spacecraft, a bubble of warped spacetime is formed around the craft so that it could cross the vast distances between stars in a matter of weeks.

Mark writes that it is heartening to know that, besides the critisim of other physicists who do not believe in the success of White’s idea, the government has spent $50,000 anyway to explore this possibility to fullfill the dream of interstellar travel. A dream that is shared by a suprising number of people who hold academic conferences on this topic and found organizations like the 100 Year Starship project, the Tau Zero Foundation and Icarus Interstellar, that seek to lay the groundwork for an unmanned interstellar mission that could be launched by the end of the century. This would be helpful to explore the slew of earthlike planets habitable for humans discovered by astronomers over the recent years.

With traditional technology probes would take thousands of years to reach planets in other solar systems. As an example the article mentions NASA’s Voyager 1, traveling  at 38,610 miles per hour,  that has left our solar system in 2012 after completing it’s primary mission to investigete the Jupiter, Saturn and their moons. With that speed it would take 70,000 years until Voyager 1 reaches any of the nearby stars that might harbor habitable planets.

As an alternative to the warp-drive, the article referes to a mission prooposed by Icarus Interstellar using nuclear fusion power for propulsion which is considered as a more realistic approach by many interstellar enthusiasts. Used properly it would allow speeds thousand of times faster than the Voyager 1. But the technology is not ready yet as researchers tried the last 50 years to use it in a power plant without success. Also the huge amount of fuel requiered for traveling these vast distances presents researchers with a big problem which is aggrevated by the heavy shielding, needed to protect the spacecraft against stardust collisions at high speeds, and decelerating from these high speeds. With regard to these enormous difficulties Mark tries to explain the paradox first noted by physiscist Enrico Fermi in 1950 that even if intelligent life in universe is common, extraterrestrials perhaps never visited earth because it is so hard to get here.

The article finishes with an argument by advocates of interstellar travel which states that planetary catastrophes threaten the long term survival of the human race, thus we must find a solution despite the difficulties. With this argument Mark comes back to the Star Trek analogy and states that we need to adopt the motto of starship Enterprise: “to boldly go where no man has gone bewfore”.

The Process of writing my bachelor thesis

What I liked

For my thesis I read a lot of related work and I actually enjoyed doing that. Getting to know the background information, the different approaches and application areas was really interesting.  I also had an easy time describing what I have done since I was motivated by the thought that this is the main contribution of my work. And with each finished chapter I could feel the shrinking workload which is always a good feeling.

The Difficulties

In the retrospective the main difficulty for me probably was the lack of a precisely defined goal for my thesis. I had a rough idea and thought if I continue working, the goal would get more precise step by step. Therefore, I often lost focus during the literature search, also because I enjoy reading related work as I mentioned above.

Getting lost in related work leads to my next problem, balancing the amount of related work in my thesis. On the one hand I might add too much unnecessary side information on the other hand I might miss to explain the building blocks to understand what I have done or why I have done it. Also clearly stating the reasons for th choice of my used approaches was hard for me.

Furthermore, I was unsure when to use references, especially concerning more basic information that can be read in any good textbook of the according research area. Additionally, struggling with the sentence structure and choosing the right expressions took ages sometimes and still does in my daily writing.

The Result

All in all, I think the final thesis turned out ok from the writing perspective thanks to the feedback of my advisor. Although the high level coherence between the chapters could have been better. I am more unsatisfied with the overall working process I had for my thesis because it lacked a precise goal for a long time. As a result, my first structural draft of my thesis came up pretty late in the process which increased the pressure in the end. That is the main reason I take part in this seminar, to improve the process of my scientific work in order to reduce the stress during writing.

Summery of my bachelor thesis

Introduction

To increase the safety and comfort in the road traffic the automotive industry is developing increasingly complex driver assistance systems (DAS) up to full autonomous cars. These systems use the data of multiple sensors and fuse them to an abstract model of the car surroundings to decide if and how the computer should actively intervene in the controls of the car to help the human driver. Because of this direct intervention these Systems need to be tested extensively to ensure the safety of all traffic participants.

Traditionally DAS-tests include Hardware and Software tests using a virtual car and environment models. Also tests under real conditions using the real car on test tracks or in real traffic are mandatory to take account of the unavoidable error made in the virtual model of the car dynamics. Unlike the virtual tests the real ones require a tremendous effort and thus cannot be executed as often as they might be needed.

Recently a new test method called Vehicle-in-the-Loop (ViL) was proposed to reduce the effort of testing with the real car. The real sensor data for perceiving the environment is replaced by a virtual model. This minimizes the test setup to the real car on a flat empty space without requiring any physical traffic dummies. The virtual model is visualized on an optical-see-through Head mounted display (HMD) that offers head tracking to give the test driver an immersive view on the virtual environment. This helps to understand the behavior of the tested DAS.

However, the optical-see-through HMD limits the degree of immersion by a small field of view and only transparent overlays. Therefore, my goal was to fuse the data of a roof-mounted 360°-Camera and -Laserscanner to create a textured mesh that can be used for a video-see-through ViL-setup and thus increase the immersion.

Fusion & Visualization

One 360° laser scan is fused with a temporal corresponding frame of the 360° camera. To correctly align the 3D information in the laserscan with the texture information in the camera image, the 6 DOF transformation between these two sensors need to be determined by a calibration procedure. Furthermore, a surface is generated on which the camera image can be mapped as a texture for an immersive visualization because the raw laserscanner data is only a point cloud.

Calibration
Besides the distance information the laserscanner also measures the strength of the reflection. Tthe statistical dependency between these reflectance values and the camera image pixel intensities is exploited by the used calibration algorithm to find the 6-DOF sensor transformation without the need of an artifical calibration target. By using an initial guess of the transformation to project the pointcloud onto the camera image, the algorithm iteratively improves the guess by maximizing the mutual information between the reflectance values and the pixel intensities the points get projected onto.

Visualization
Due to the rotational gathering strategy of the laserscanner the point cloud can be structured into rows and columns with missing points where no depth value could be measured. The mesh is generated by connecting two neighbored points in a row with a neighbored point in the next column to form a triangle, repeating for all points.

Evaluation
The presented method was evaluated by qualitatively analyzing the results when applied to test datasets gathered by an experimental vehicle. Furthermore, a reference camera image from a passenger’s perspective was compared to the visualization result of the same perspective by computing a correlation metric.

The generated textured mesh generally allows a good spatial understanding of the environment because of the combined 3D- and color information. But due to missing range measurements and the simple meshing technique it contains many holes that drastically decreases the perceived quality. This gets reinforced by an 41% increment of the correlation value when the missing depth values were filled by a predefined value.

Also alignment artifacts occur on objects that are moving relative to the sensor system because the temporal correlating camera and scanner data is not acquired at exactly the same time. Therefore, the perceived position of the object differs between the sensors leading to the misalignment.

Conclusion

I presented a method to generate a textured mesh of the surroundings of a car that is intended to be used in a video-see-through ViL setup. The calibration of the 360°-camera to the 360°-laserscanner is easy to execute because it does not require an artificial calibration target and it is accurate when used with a meaningful initial guess.

The visualization lags with 5Hz behind the 10Hz update rate of the laserscanner and contains misaligned textures of moving objects due to small differences in sensor data acquisition times. Also the holes and flickering in the mesh lead to an overall poor quality for an immersive visualization. Therefore, the developed system should not be used in an HMD in the current state.

Future work could target the aforementioned problems or extend the method by registering successive meshes to generate an environmental map.