Adding a receding horizon to Locally Weighted Regression for learning robot control

There have been notable advances in learning to control complex robotic systems using methods such as Locally Weighted Regression (LWR). In this paper we explore some potential limits of LWR for robotic applications, particularly investigating its application to systems with a long horizon of temporal dependence. We define the horizon of temporal dependence as the delay from a control input to a desired change in output. LWR alone cannot be used in a temporally dependent system to find meaningful control values from only the current state variables and output, as the relationship between the input and the current state is under-constrained. By introducing a receding horizon of the future output states of the system, we show that sufficient constraint is applied to learn good solutions through LWR. The new method, Receding Horizon Locally Weighted Regression (RH-LWR), is demonstrated through one-shot learning on a real Series Elastic Actuator controlling a pendulum.

United States still no closer to database legislation

Last year European Intellectual Property Review published an article comparing the latest version of the proposed US database legislation, the Collections of Information Antipiracy Bill with the UK's Copyright and Rights in Database Regulations 1997. Subsequently a new US Bill, the Consumer and Investor Access to Information Act has emerged, the Antipiracy Bill has been amended and much debate has occurred, but the US seems no closer to enacting database legislation. This article briefly outlines the background to the US legislative efforts, examines the two Bills and draws some comparisons with the UK Regulations. A study of the US Bills clearly demonstrates the starkly divided opinion on database protection held by the Bills' proponents and the principal lobby groups driving the legislative efforts: the Antipiracy Bill is very protective of database producers' interests, whereas the Access Bill is heavily user-oriented. If the US experience is any indication there will be a long horizon involved in achieving any consensus on international harmonisation of this difficult area.

Jackknife for bias reduction in predictive regressions

One of the fundamental econometric models in finance is predictive regression. The standard least squares method produces biased coefficient estimates when the regressor is persistent and its innovations are correlated with those of the dependent variable. This article proposes a general and convenient method based on the jackknife technique to tackle the estimation problem. The proposed method reduces the bias for both single- and multiple-regressor models and for both short- and long-horizon regressions. The effectiveness of the proposed method is demonstrated by simulations. An empirical application to equity premium prediction using the dividend yield and the short rate highlights the differences between the results by the standard approach and those by the bias-reduced estimator. The significant predictive variables under the ordinary least squares become insignificant after adjusting for the finite-sample bias. These discrepancies suggest that bias reduction in predictive regressions is important in practical applications.

In conversation with the Honourable Ian Callinan AC

When, in 1977, the Australian electorate provided a double majority to effect a change of section 72 of the Commonwealth Constitution requiring judges of the High Court of Australia to retire at the age of 70 years old, I doubt we understood the continuing capacity of these esteemed members of the judiciary. For the opportunity to sit and talk with Ian Callinan AC who, in compliance with that amendment, retired from the High Court in September 2007, I needed to wait until he returned from The Hague where he was sitting as a Judge ad hoc on the International Court of Justice. Although a native of Casino, New South Wales, Mr Callinan is regarded as a Queenslander. Indeed, he grew up in Brisbane, finished high school at Brisbane Grammar and graduated in law at The University of Queensland. Appointed in 1978 as a Queen’s Counsel, Mr Callinan enjoyed this period of his legal career and we discussed an aspect of the Christopher Skase case, which reinforced my belief that Mr Callinan is an incredibly skilful advocate. On 14 September 1998, ABC Four Corners broadcasted the views of some prominent Australians on the appointment of Mr Callinan to the High Court. In assessing the type of person Mr Callinan is, Tony Morris QC said: “Ian Callinan isn't a coward”, while former Commonwealth Attorney-General, Michael Lavarch, said: “He was regarded as an absolutely outstanding criminal lawyer within the Queensland legal profession, I mean really a top-notch advocate”. I was not interested in raising any of the controversial issues that Mr Callinan has encountered as an advocate in high profile matters. I wanted to know how he felt about his time on the High Court, what his thoughts are on the operation of the High Court, the IP cases he decided, the real life issues that he feels impact on counsel who are appearing before the High Court and the people he regarded as role models. During our conversation, Mr Callinan laughed often and when he did his eyes lit up, revealing his passion for life. He is an incredibly genuine Australian who loved his time as a barrister, enjoyed his role on the High Court, enjoys his current job as mediator, loves writing novels, has a great desire for continual improvement in the quality of legal education and legal advocacy and sees a need for change in IP law. When I asked: “So, what might the future hold for you?”, he laughed and said: “Well, at my age I don’t have a long horizon time”. I said: “Just enjoy the journey?”, to which Mr Callinan responded: “Exactly”.

The value of tail risk hedging in defined contribution plans: What does history tell us

Hedging against tail events in equity markets has been forcefully advocated in the aftermath of recent global financial crisis. Whether this is beneficial to long horizon investors like employees enrolled in defined contribution (DC) plans, however, has been subject to criticism. We conduct historical simulation since 1928 to examine the effectiveness of active and passive tail risk hedging using out of money put options for hypothetical equity portfolios of DC plan participants with 20 years to retirement. Our findings show that the cost of tail hedging exceeds the benefits for a majority of the plan participants during the sample period. However, for a significant number of simulations, hedging result in superior outcomes relative to an unhedged position. Active tail hedging is more effective when employees confront several panic-driven periods characterized by short and sharp market swings in the equity markets over the investment horizon. Passive hedging, on the other hand, proves beneficial when they encounter an extremely rare event like the Great Depression as equity markets go into deep and prolonged decline.

Soil carbon saturation : implications for measurable carbon pool dynamics in long-term incubations

The efficiency of agricultural management practices to store SOC depends on C input level and how far a soil is from its saturation level (i.e. saturation deficit). The C Saturation hypothesis suggests an ultimate soil C stabilization capacity defined by four SOM pools capable of C saturation: (1) non-protected, (2) physically protected, (3) chemically protected and (4) biochemically protected. We tested if C saturation deficit and the amount of added C influenced SOC storage in measurable soil fractions corresponding to the conceptual chemical, physical, biochemical, and non-protected C pools. We added two levels of C-13- labeled residue to soil samples from seven agricultural sites that were either closer to (i.e., A-horizon) or further from (i.e., C-horizon) their C saturation level and incubated them for 2.5 years. Residue-derived C stabilization was, in most sites, directly related to C saturation deficit but mechanisms of C stabilization differed between the chemically and biochemically protected pools. The physically protected C pool showed a varied effect of C saturation deficit on C-13 stabilization, due to opposite behavior of the POM and mineral fractions. We found distinct behavior between unaggregated and aggregated mineral-associated fractions emphasizing the mechanistic difference between the chemically and physically protected C-pools. To accurately predict SOC dynamics and stabilization, C Saturation of soil C pools, particularly the chemically and biochemically protected pools, should be considered. (C) 2008 Elsevier Ltd. All rights reserved.

Soil carbon saturation : evaluation and corroboration by long-term incubations

Although current assessments of agricultural management practices on soil organic C (SOC) dynamics are usually conducted without any explicit consideration of limits to soil C storage, it has been hypothesized that the SOC pool has an upper, or saturation limit with respect to C input levels at steady state. Agricultural management practices that increase C input levels over time produce a new equilibrium soil C content. However, multiple C input level treatments that produce no increase in SOC stocks at equilibrium show that soils have become saturated with respect to C inputs. SOC storage of added C input is a function of how far a soil is from saturation level (saturation deficit) as well as C input level. We tested experimentally if C saturation deficit and varying C input levels influenced soil C stabilization of added C-13 in soils varying in SOC content and physiochemical characteristics. We incubated for 2.5 years soil samples from seven agricultural sites that were closer to (i.e., A-horizon) or further from (i.e., C-horizon) their C saturation limit. At the initiation of the incubations, samples received low or high C input levels of 13 C-labeled wheat straw. We also tested the effect of Ca addition and residue quality on a subset of these soils. We hypothesized that the proportion of C stabilized would be greater in samples with larger C Saturation deficits (i.e., the C- versus A-horizon samples) and that the relative stabilization efficiency (i.e., Delta SCC/Delta C input) would decrease as C input level increased. We found that C saturation deficit influenced the stabilization of added residue at six out of the seven sites and C addition level affected the stabilization of added residue in four sites, corroborating both hypotheses. Increasing Ca availability or decreasing residue quality had no effect on the stabilization of added residue. The amount of new C stabilized was significantly related to C saturation deficit, supporting the hypothesis that C saturation influenced C stabilization at all our sites. Our results suggest that soils with low C contents and degraded lands may have the greatest potential and efficiency to store added C because they are further from their saturation level. (c) 2008 Elsevier Ltd. All rights reserved.

Robust scale initialization for long-range stereo visual odometry

Achieving a robust, accurately scaled pose estimate in long-range stereo presents significant challenges. For large scene depths, triangulation from a single stereo pair is inadequate and noisy. Additionally, vibration and flexible rigs in airborne applications mean accurate calibrations are often compromised. This paper presents a technique for accurately initializing a long-range stereo VO algorithm at large scene depth, with accurate scale, without explicitly computing structure from rigidly fixed camera pairs. By performing a monocular pose estimate over a window of frames from a single camera, followed by adding the secondary camera frames in a modified bundle adjustment, an accurate, metrically scaled pose estimate can be found. To achieve this the scale of the stereo pair is included in the optimization as an additional parameter. Results are presented both on simulated and field gathered data from a fixed-wing UAV flying at significant altitude, where the epipolar geometry is inaccurate due to structural deformation and triangulation from a single pair is insufficient. Comparisons are made with more conventional VO techniques where the scale is not explicitly optimized, and demonstrated over repeated trials to indicate robustness.